Why it matters
@bscholl invites viewers to a shop-floor walkthrough with Boom Supersonic’s R&D shop manager showing how cold-section engine vanes are manufactured from raw materials.
Key details
- Boom Supersonic launched a 'Build Supersonic' video series; Episode 1 (posted May 11, 2026) spotlights 'Symphony' hardware and rapid iterative testing—'how we break things, learn, and iterate to move fast'—and the post encourages people to join the team.
Brief
@bscholl invites viewers to walk the shop floor with Boom Supersonic’s R&D shop manager to see how cold-section engine vanes are made from raw materials. Boom launched the 'Build Supersonic' video series; Episode 1 highlights Symphony hardware and their rapid iterate-and-test approach ('how we break things, learn, and iterate to move fast') and asks what viewers want to see next while encouraging new hires.
By @bscholl
Why it matters
A 1 GW continuous data center emits 24 billion watt‑hours (24 GWh) of heat per day — comparable in magnitude to the Trinity Test’s ~27 GWh total heat release, but the Trinity blast released that energy in ~1 second (≈86,000× more concentrated).
Key details
- Reporting waste heat in “nuclear bombs” is misleading: the article models a proposed site with 16 GW thermal (9 GW data center + 7 GW gas turbines) → 16 GW × 24 h = 384 GWh/day; using 1 GWh ≈ 0.86 kilotons TNT gives ~16.7 GWh per bomb (~14.3 kt, similar to Little Boy), an alarming but concentration‑incommensurate comparison.
- Local and meteorological heat flows dwarf or contextualize data‑center heat: Washington DC’s total human energy use ≈65 GWh/day (~28 GWh electricity + 37 GWh transport/other) ≈ 3.7 ‘nukes’/day, while incoming solar ≈750–1000 GWh/day (≈43–57 ‘nukes’/day) and 1 mm of condensation over DC ≈108 GWh (~6 ‘nukes’).
- A more useful metric is local temperature change: using an estimate of ~0.025 °C per W/m² of human energy use implies DC’s human energy adds ≈0.4 °C, and the largest proposed data center (dissipating ~16 GW) could raise very local air temperature by ~2.4 °C (~4.3 °F).
By Andy Masley
Why it matters
Varick (author @vasuman) insists “AI is a services game”: implementers must refactor business processes for an AI-native reality and build custom agents rather than relying on tool adoption alone.
Key details
- Varick claims a perfect track record: 100% of their clients have reached production and 100% returned for a second project; OpenAI announced the OpenAI Deployment Company on 2026-05-11, majority-owned by OpenAI and backed by 19 investment firms/consultancies/system integrators.
- Author warns against vendor lock-in risks—price hikes, model quantization/regressions, rate limits, and migration costs—and is recruiting Engineers, FDEs, and Consultants, inviting top 1% talent to join Varick Agents.
Brief
Varick argues AI success requires implementation services: teams that analyze and refactor processes and build custom agents, not just selling tools. He touts Varick’s record (100% production, 100% repeat clients), notes OpenAI launched an OpenAI Deployment Company on 2026-05-11 backed by 19 partners, warns of vendor lock-in risks, and is hiring Engineers, FDEs, and Consultants.
By @vasuman
Why it matters
@ancerj asserts "One way or another Taiwan will be reunified with China," arguing the next decade is about timing and terms and moving key technology assets out while buying time with regional military build-up to deny a Chinese Monroe Doctrine.
Key details
- On May 8 the KMT and Taiwan People's Party passed a NT$780 billion (~$25 billion) defense bill (vote 59-0; DPP abstained), cutting President Lai Ching-te's proposed $40 billion budget to roughly two-thirds and eliminating all domestic procurement: 210,000 military drones, sea-attack drone programs, the Chiang Kung anti-ballistic missile (T‑Dome backbone), and NT$64 billion for Taiwan–US joint R&D.
- Political and strategic fallout: DPP lawmaker Chen Kuan-ting warned reliance on imported US weapons risks ammunition and sustainment shortfalls if blockaded; KMT chair Cheng Li-wun met Xi Jinping in April 2026 weeks before the cuts; Sen. Roger Wicker expressed disappointment and Adm. Paparo has warned the US cannot want Taiwan's defense more than Taiwan itself—author argues the opposition handed Beijing a capability gap and proof of "dysfunctional" democracy.
Brief
@ancerj argues "One way or another Taiwan will be reunified with China," and that the next decade will determine timing/terms while Taiwan moves key tech assets out and builds regional military capacity to deter a Chinese Monroe Doctrine. On May 8 the KMT and TPP passed a NT$780 billion (~$25bn) defense bill (59-0; DPP abstained) stripping domestic programs—210,000 drones, sea-attack drones, Chiang Kung ABM and NT$64bn for Taiwan–US R&D—creating capability gaps and political controversy.
By @ancerj
Why it matters
Author @cryptopunk7213 claims the Cerebras IPO was 20x oversubscribed and that demand led the company to raise an extra $1.3B since last week’s IPO spec, signaling investor hunger for an NVIDIA alternative.
Key details
- The author asserts inference will be 10-50x the value of training and highlights Cerebras’ SRAM chips as specialized for low-latency inference, noting Codex Spark and Amazon Bedrock run on their chips and provide distribution.
- Reuters/Bloomberg reporting: IPO price range was increased to $150–$160 (from an original $115–$125), shares rose to 30 million from 28 million, potential raise ≈ $4.8B at $160, indications of interest > $10B, pricing set for May 13, 2026.
Brief
Cerebras’ IPO surge: the author claims a 20x oversubscription and an extra $1.3B raise, arguing strong investor appetite for an NVIDIA alternative if Cerebras captures even 10% market share. He emphasizes Cerebras’ SRAM chips for low-latency inference (Codex Spark, Amazon Bedrock) and notes Reuters reports the range rose to $150–$160 with pricing on May 13, 2026.
By @cryptopunk7213
Why it matters
Introduces the 'discriminative factorization' to evaluate query-set quality for black-box model-level classification; under this framework the probability of chance-level classification decays exponentially with the query budget.
Key details
- Empirical validation on three auditing tasks (Helm, Ohata, Priebe; arXiv 2026-05-08) shows estimated factorization parameters predict the observed performance-decay rate, and query sets chosen by the estimated discriminative field reproduce the empirical ordering of oracle query sets.
Brief
Discriminative factorization is proposed to quantify and distinguish high- versus low-quality query sets for black-box model-level classification. The framework yields a theoretical result: the probability of chance-level classification decays exponentially with query budget. On three auditing tasks the authors show estimated parameters track empirical decay and enable query selection that mirrors oracle ordering.
Authors: Hayden Helm, Merrick Ohata, Carey Priebe
Why it matters
CUTS-GPR introduces an extremely fast kernel matrix–vector product that attains near-linear or linear scaling with training size N and low-order polynomial scaling with dimensionality D by combining an additive kernel with an incomplete grid structure.
Key details
- Authors demonstrate scalability with benchmarks involving billions of data points and thousands of dimensions; a full Gaussian process regression (including hyperparameter optimization) was completed in hours for N = 447,265 and D = 24, enabling Bayesian modeling of high-dimensional potential energy surfaces.
Brief
CUTS-GPR targets the computational bottleneck of Gaussian process regression in high-dimensional problems by exploiting an additive kernel on an incomplete grid to reveal structure that yields an extremely fast kernel matrix–vector product. The method shows near-linear (or linear) scaling in N and low-order polynomial scaling in D, with benchmarks on billions of points and thousands of dimensions and a full GPR + hyperparameter run in hours for N=447,265, D=24, enabling Bayesian modeling of high-dimensional potential energy surfaces in computational chemistry.
Authors: Mads Greisen Højlund, August Smart Lykke-Møller, Henry Moss...
Why it matters
Zou, Poeppel and Ding (Nature Neuroscience) — highlighted by @ValerioCapraro on 2026-05-11 — show the human brain predicts words but prediction precision is constrained by linguistic structure.
Key details
- When a word continues the current phrase, brain activity tracks word surprisal in a way resembling LLMs; when a word crosses a major phrase boundary, the match with LLM-style prediction weakens.
- The authors and poster assert this challenges the view that humans are mere next-token predictors: the brain asks not just "What is the next word?" but also "What structure am I currently building?"
Brief
Zou, Poeppel and Ding's Nature Neuroscience paper, highlighted by @ValerioCapraro on 2026-05-11, reports that human word prediction varies with syntactic structure: surprisal tracking aligns with LLM-like next-word prediction within phrases but breaks down across major phrase boundaries, arguing human language processing involves structural tracking beyond next-token prediction.
By @ValerioCapraro
Why it matters
DR-ME, proposed by Houssam Zenati and Arthur Gretton (arXiv 2026-05-08), is the first semiparametrically efficient finite-location test that returns interpretable causal-discrepancy coordinates for distributional treatment effects rather than only a global rejection.
Key details
- The method constructs orthogonal, doubly robust kernel features from observational data whose centered oracle form is the canonical gradient of the finite witness; for fixed locations the test is chi-square calibrated under the null and has noncentral chi-square local power, employing covariance whitening that optimizes local signal-to-noise.
- DR-ME uses a principled location-learning criterion with sample splitting to preserve post-selection validity; experiments report near-nominal Type-I error, competitive power versus global doubly robust kernel tests, and interpretable learned locations in a semi-synthetic medical-imaging study.
Brief
Distributional treatment effects that leave means unchanged are targeted by DR-ME, a semiparametrically efficient finite-location test (Zenati & Gretton, arXiv:2605.08034v1, 2026-05-08). From observational data it derives orthogonal doubly robust kernel features whose centered oracle is the canonical gradient; for fixed locations the test is chi-square calibrated and has noncentral chi-square local power with covariance whitening optimizing local SNR. Sample splitting preserves post-selection validity; experiments show near-nominal Type-I error and competitive power, with learned locations that localize effects in a semi-synthetic medical-imaging study.
Authors: Houssam Zenati, Arthur Gretton
Why it matters
Bohlinger et al. (arXiv 2026-05-08) evaluate an actuated 1-DOF spine in the sagittal plane on MAB Robotics' Silver Badger in MuJoCo, finding that the spine provides increased agility and enables the robot to overcome higher stairs, steeper slopes, taller hurdles, and smaller passages versus the non-spined configuration.
Key details
- The empirical simulation study covers multiple tasks—high-speed running, stair climbing, high-angle slope climbing, hurdling, and crawling—demonstrating consistent performance gains from adding a single-DOF spinal actuator for learned agile quadruped locomotion.
Brief
The paper investigates whether a single-DOF actuated sagittal spine improves learned agile locomotion for a quadruped. Using MuJoCo simulations of the Silver Badger robot, the authors evaluate high-speed running, stair and high-angle slope climbing, hurdling, and crawling. Results show the spine increases agility and enables traversing higher stairs, steeper slopes, taller obstacles, and narrower passages, suggesting spine actuation is a promising design extension for agile robots.
Authors: Nico Bohlinger, Piotr Kicki, Davide Tateo...
Why it matters
@swyx posted what he believes are the first public photos of Cognition’s "Cog House" and, as an advisor, says the company will be worth $100B by the end of 2026 (his opinion).
Key details
- Scott Wu, Cognition co‑founder, is a competitive‑programming prodigy: three IOI gold medals, US national middle‑school math champion, widely regarded as America’s top IOI gold‑medalist and coach.
- Cognition was founded in November 2023 when Wu was 26; it shipped AI engineer "Devin" in March 2024, which after criticism helped the company reach a $445M revenue run rate within 18 months, with usage doubling every eight weeks and customers including the US Army, Goldman Sachs, and Mercedes‑Benz; the company is raising at about a $25B valuation.
Brief
@swyx shared purportedly the first public photos of Cognition’s secretive "Cog House" and praises the company’s trajectory, predicting a $100B valuation by end of 2026. The post highlights a Colossus profile of co‑founder Scott Wu—a three‑time IOI gold medalist—who founded Cognition in Nov 2023, launched Devin in Mar 2024, and grew to a $445M run rate with major customers and a ~ $25B raise.
By @swyx
Why it matters
Introduces a Householder-aligned permutation test that applies a single Householder reflection to align mean directions of two token-vector clouds before nonparametric permutation testing, isolating dispersion (semantic breadth) from directional differences in contextualized embeddings.
Key details
- On evaluated cases the alignment reduced Type‑I error by 32.5% while preserving sensitivity to true breadth differences (compared with naive tests that confound direction and dispersion).
- Provides a GPU-oriented, batched implementation that achieves a 23× speedup over the CPU baseline; paper by Yo Ehara posted to arXiv 2026-05-08 and accepted to ACL 2026 Main Conference.
Brief
Yo Ehara proposes a Householder-aligned permutation test to more accurately compare word semantic breadth from contextualized token embeddings. By reflecting one token cloud to align mean directions before permutation testing, the method prevents directional differences from masquerading as dispersion effects. Empirically it cut Type‑I error by 32.5% and, with a GPU-batched implementation, ran 23× faster than a CPU baseline; accepted to ACL 2026.
Authors: Yo Ehara
Why it matters
On 2026-05-10 23:34:48+00:00 Substack (mg1.substack.com) recorded that user 'Tpayneful' liked the post titled "The Factory's Quote is the Product."
Key details
- The notification references author Spencer Burleigh (©2026) and includes a San Francisco contact address (548 Market Street PMB 72296, San Francisco, CA 94104) and a displayed figure "681" (likely a reader/engagement metric).
Brief
A Substack notification shows that user Tpayneful liked Spencer Burleigh’s post "The Factory's Quote is the Product" on mg1.substack.com, with a published timestamp of 2026-05-10 23:34:48+00:00 (created 2026-03-11, updated 2026-04-04). The message includes author metadata (©2026), profile images, action links and a displayed metric "681," plus a San Francisco mailing address.
By Tpayneful
Why it matters
On 2026-05-10 23:34:31+00:00 Substack user 'Tpayneful' liked Spencer Burleigh’s post titled "Oil’s Financials are Falling Apart" on mg1.substack.com.
Key details
- Article metadata shows it was created 2026-03-11 12:44 and last updated 2026-04-04 00:00; the content is a Substack like/notification email linking to the post and Burleigh’s profile (Spencer Burleigh, 548 Market Street PMB 72296).
Brief
Tpayneful liked Spencer Burleigh’s Substack post "Oil’s Financials are Falling Apart" (notification timestamp 2026-05-10 23:34:31+00:00). The message is a Substack email/like notification linking to the original post (created 2026-03-11 12:44; last updated 2026-04-04 00:00) and Burleigh’s author profile.
By Tpayneful
Why it matters
On 2026-05-10 23:38:33+00:00 Substack user 'Tpayneful' liked the post titled 'Snap - When your Core Competency is a Loss Leader' hosted on mg1.substack.com.
Key details
- Document metadata shows Created: 2026-03-11 12:44 and Last Updated: 2026-04-04 00:00; the page also displays © 2026 Spencer Burleigh and the address 548 Market Street PMB 72296, San Francisco, CA 94104.
- The submitted content is a Substack notification/UI page with images and links but contains no substantive article body or analysis text beyond the like/notification details.
Brief
A Substack notification records that user 'Tpayneful' liked the post 'Snap - When your Core Competency is a Loss Leader' on mg1.substack.com (liked at 2026-05-10 23:38:33 UTC). Metadata lists Created 2026-03-11 12:44 and Last Updated 2026-04-04; the page includes © 2026 Spencer Burleigh and a San Francisco mailing address but no article body.
By Tpayneful
Why it matters
On 2026-05-11 Al Mayadeen English shared Professor Mohammad Marandi's claim that UAE President Sheikh Mohamed bin Zayed Al Nahyan (MBZ) has taken his alignment with Israel "far too far," calling the UAE "very destructive for the Islamic world, for the Arab world, and for humanity."
Key details
- Marandi asserted the UAE's regional policies are "in line with Zionist interests in the Horn of Africa, in the Arabian Peninsula and in the Persian Gulf," and @ianmiles responded to the post with the blunt rebuttal: "Oh fuck off."
Brief
Professor Mohammad Marandi, quoted by Al Mayadeen English on 2026-05-11, accused UAE President Sheikh Mohamed bin Zayed Al Nahyan of moving "far too far" toward Israel and described the UAE as "very destructive for the Islamic world, for the Arab world, and for humanity," alleging alignment with "Zionist interests" across the Horn of Africa, Arabian Peninsula and Persian Gulf; the post drew a curt reply from @ianmiles: "Oh fuck off."
By @ianmiles
Why it matters
Mallis, Wang, Karadeniz, Ricci, Kacem, and Aouada (arXiv 2026-05-08) introduce CADTestBench and CADTests: the first test-based benchmark for Text-to-CAD where CADTests are executable software tests that verify geometric and topological requirements of generated CAD models.
Key details
- Using CADTestBench the authors benchmark recent Text-to-CAD methods and show CADTests can also guide model generation, producing simple baselines that surpass current methods; code and data are published on GitHub and as a Hugging Face dataset.
Brief
Text-to-CAD evaluation is framed as automated testing in this work: Mallis et al. propose CADTestBench and CADTests, executable checks that validate geometric and topological constraints of generated CAD models. They benchmark recent Text-to-CAD systems on CADTestBench and demonstrate that using CADTests to guide generation yields simple baselines that outperform prior methods; code and datasets are open-sourced.
Authors: Dimitrios Mallis, Marco Wang, Ahmet Serdar Karadeniz...
Why it matters
Vibe-Trading has shipped daily updates for an entire month since its open-source release and surpassed 6,000 GitHub stars (post published 2026-05-11).
Key details
- The framework couples AI agents with quantitative trading tools, directly integrates with data sources Tushare and AKShare, and supports Chinese A‑shares, crypto, and international markets with exports to TradingView, 通达信, and MetaTrader.
- Key features include brokerage-statement upload for instant analytics and a "Shadow Account" performance view; persistent-memory AI agents that self-improve and support 13+ AI models; and enterprise-ready security with API authentication, secure execution environments, and Docker deployment.
Brief
Vibe-Trading is an open-source AI + quant trading framework that hit 6k+ GitHub stars and has shipped daily updates for a month (published 2026-05-11). It connects to Tushare and AKShare, supports A‑shares/crypto/global markets, exports to TradingView/通达信/MetaTrader, and offers brokerage analytics, a Shadow Account, persistent-memory agents (13+ models), and Docker-ready security for AI+Human workflows.
By @huang_chao4969
Why it matters
Citrini (note posted ~7 days before the digest) reports Elon Musk saying actuators make up 56% of the Bill of Materials for Tesla's Optimus robot; that note recorded 84 likes, 8 comments, and 8 restacks.
Key details
- Sam Bowman (note posted ~6 days before the digest) highlights emerging hair‑loss treatments that may be effective without travel; the post received 73 likes, 4 comments, and 6 restacks.
- Internal Tech Emails (note posted ~5 days before the digest) republishes messages including a ‘Sam Altman texts Mira Murati…’ item; it drew the largest engagement in the digest with 116 likes, 9 comments, and 13 restacks.
Brief
Three Substack notes (Citrini, Sam Bowman, Internal Tech Emails) featured in the May 11, 2026 digest. Citrini cites Elon Musk claiming actuators account for 56% of Optimus’s BOM. Sam Bowman summarizes promising hair‑loss treatments. Internal Tech Emails republishes internal messages (including Sam Altman–Mira Murati texts). Engagement counts were 84/73/116 likes and 8/4/9 comments respectively.
By Substack
Why it matters
User Tpayneful liked the Substack post "American Words - Still Everywhere" on 2026-05-10 23:46:16+00:00; the post metadata shows it was created 2026-03-11 (12:44) and last updated 2026-04-04 (00:00).
Key details
- The notification is from mg1.substack.com / Spencer Burleigh (© 2026) and includes publisher contact/address 548 Market Street PMB 72296, San Francisco, CA 94104, with links to the post and user profile.
Brief
American Words - Still Everywhere is a Substack post that received a like from user Tpayneful on 2026-05-10 23:46:16 UTC. The post was created 2026-03-11 (12:44) and last updated 2026-04-04 (00:00); the notification originates from mg1.substack.com under Spencer Burleigh (©2026) and contains links and UI metadata but not the article body.
By Tpayneful
Why it matters
Proxy3D builds compact 3D proxy representations from only video frames using semantic and geometric encoders plus semantic-aware clustering to produce scene proxies in 3D space.
Key details
- The authors curated the SpaceSpan dataset and use multi-stage training to align these proxies with vision-language models; the approach yields competitive or state-of-the-art performance on 3D visual question answering, visual grounding, and spatial intelligence benchmarks while using shorter vision sequences.
- Paper by Jerry Jiang, Haowen Sun, Denis Gudovskiy et al., posted to arXiv 2026-05-08 and accepted to CVPR 2026; project page: https://wzzheng.net/Proxy3D
Brief
Proxy3D proposes compact 3D proxy representations extracted from video frames via semantic and geometric encoders and semantic-aware clustering. The authors curate the SpaceSpan dataset and apply multi-stage training to align proxies with vision-language models; on 3D visual question answering, visual grounding, and spatial intelligence benchmarks the method achieves competitive or state-of-the-art results while using shorter vision sequences. (Abstract-only summary.)
Authors: Jerry Jiang, Haowen Sun, Denis Gudovskiy...
Why it matters
On 2026-05-10 23:37:13+00:00 Substack user 'Tpayneful' liked the post 'Founder's Guide to Building a Second Brain' (notification contains links to the post and the liker’s profile).
Key details
- The notification originates from Spencer Burleigh's Substack (© 2026 Spencer Burleigh) and includes action links ('View profile', 'Mute post') and a displayed numeric '843' on the page.
- Document metadata shows Created: 2026-03-11-12-44 and Last Updated: 2026-04-04-00-00; the footer lists 548 Market Street PMB 72296, San Francisco, CA 94104.
Brief
Tpayneful liked the Substack post 'Founder's Guide to Building a Second Brain' on 2026-05-10 23:37:13+00:00. The notification email/page is from Spencer Burleigh's Substack (© 2026 Spencer Burleigh), includes links to the post and the liker’s profile plus controls (View profile, Mute post), and shows metadata Created 2026-03-11 12:44 and Last Updated 2026-04-04.
By Tpayneful
Why it matters
Flow-OPD applies on-policy distillation to Flow Matching text-to-image models with a two-stage alignment: (1) single-reward GRPO fine-tuning to produce domain-specialized teacher experts; (2) Flow-based Cold-Start and a three-step student consolidation (on-policy sampling, task‑routing labeling, dense trajectory-level supervision). It also introduces Manifold Anchor Regularization (MAR) to anchor outputs to a high-quality manifold.
Key details
- Built on Stable Diffusion 3.5 Medium, Flow-OPD raises GenEval from 63 to 92 and OCR accuracy from 59 to 94, claims an overall improvement of roughly 10 points over vanilla GRPO, preserves image fidelity and human-preference alignment, and exhibits an emergent 'teacher-surpassing' effect (paper published 2026-05-08).
Brief
Flow-OPD addresses reward sparsity and gradient interference in multi-task alignment for Flow Matching text-to-image models by combining single-reward GRPO teachers with a Flow-based Cold-Start and on-policy distillation into a single student, plus Manifold Anchor Regularization to prevent aesthetic degradation. Built on Stable Diffusion 3.5 Medium, the abstract reports GenEval 63→92 and OCR 59→94; full text was not available.
Authors: Zhen Fang, Wenxuan Huang, Yu Zeng...
Why it matters
The authors derive a closed-form expression proving online kernel regression is equivalent to offline kernel regression with systematically shifted (inaccurate) target outputs, and they show that compensating for this effective target shift can provably recover the offline predictor.
Key details
- They provide both a closed-form target-correction and an iterative sequential form; empirically, online SGD with iteratively corrected targets outperforms learning with the true targets on CIFAR-10 and CORe50 in continual-learning settings (paper by Ziyan Li and Naoki Hiratani, arXiv:2605.07886v1, published 2026-05-08; 22 pages, 6 figures).
Brief
Online kernel regression: Li and Hiratani (2026) derive a closed-form expression showing online kernel regression is equivalent to offline kernel regression with shifted, inaccurate target outputs. They give a closed-form and an iterative target-correction that provably recovers the offline predictor. Experiments on CIFAR-10 and CORe50 show online SGD with corrected targets outperforms using true targets in continual learning.
Authors: Ziyan Li, Naoki Hiratani
Why it matters
On 2026-05-10 23:39:12+00:00 Substack user Tpayneful liked the post titled "Storytelling and Hope - Everywhere in Life" (mg1.substack.com).
Key details
- The notification is from a Substack newsletter associated with Spencer Burleigh (© 2026) and includes links to view the original post, the liker’s profile, and a mute option; the footer lists 548 Market Street PMB 72296, San Francisco, CA 94104.
Brief
Tpayneful liked the Substack post "Storytelling and Hope - Everywhere in Life" on 2026-05-10 23:39:12 UTC. The item is a Substack notification (mg1.substack.com) linked to Spencer Burleigh’s newsletter (© 2026), containing direct links to the post and profile, a mute control, and a San Francisco mailing address in the email footer.
By Tpayneful
Why it matters
On 2026-05-10 23:38:07+00:00 Substack user Tpayneful liked the post titled "Chicken - Inconsistent with the UK's Brand" (notification shows the like event and the post title).
Key details
- Metadata shows the item was created 2026-03-11 12:44, last updated 2026-04-04 00:00, is associated with Spencer Burleigh (© 2026) and includes a footer listing "894" and the address 548 Market Street PMB 72296, San Francisco, CA 94104.
Brief
The Substack notification records that user Tpayneful liked the post "Chicken - Inconsistent with the UK's Brand" on 2026-05-10 23:38:07 UTC. The page metadata lists creation on 2026-03-11 12:44 and last update on 2026-04-04 00:00, attributes the content to Spencer Burleigh (© 2026), and includes a footer showing "894" and a San Francisco address.
By Tpayneful
Why it matters
Anniversary refresh (reported May 11, 2026) adds a 5X earning rate on hotel spending to a popular dining-focused credit card.
Key details
- The update includes nearly $100 in limited-time travel and dining credits and enrollable rental-car elite status; enrollment is required to activate these benefits.
- The card is being promoted with a top-tier welcome offer as high as 100,000 points; AwardWallet notes full details are behind a blog link and says checking the offer won’t hurt your credit score.
Brief
A popular dining-focused credit card received an anniversary refresh (reported May 11, 2026) that adds 5X points on hotels, nearly $100 in limited-time travel and dining credits, and enrollable rental-car elite status; the promotion also advertises a welcome bonus up to 100,000 points and requires enrollment for some benefits.
By AwardWallet
Why it matters
On 2026-05-11, Gary Marcus replied to Geoffrey Hinton, explicitly denying he ever said AI systems “JUST regurgitate” and insisting that claim is false.
Key details
- Marcus concedes AI systems sometimes regurgitate and calls the evidence for that “overwhelming,” but distinguishes regurgitation from hallucinations and says he has warned about hallucinations since 2001.
- Marcus says he cannot find Hinton’s alleged quote outside Hinton’s own webpage and promises to discuss Hinton’s reply and the alleged quote in a further reply.
Brief
Gary Marcus, replying to Geoffrey Hinton on May 11, 2026, rejects Hinton’s characterization that Marcus said AI systems “JUST regurgitate,” stating he never made that claim. He accepts that models sometimes regurgitate (calling the evidence overwhelming), distinguishes regurgitation from hallucinations (which he’s warned about since 2001), and notes he cannot source Hinton’s alleged quote beyond Hinton’s webpage, promising further comment.
By @GaryMarcus
Why it matters
@0xd1namit (published 2026-05-11) says prop trading is now on Polymarket and urges the community to "can't sleep on it," crediting @BagCalls and his team for strong marketing and predicting top-tier execution; defines prop trading as firms providing their own capital after traders pass challenges.
Key details
- Funding Predicts Beta Competition (powered by Polymarket) is live: offers over 1.4m in funded accounts and 10K+ in cash prizes, gives entrants 14 days to prove they can trade, and is promoted at fundingpredicts.com (tweet/video shared by @FundingPredicts).
Brief
Prop trading on Polymarket is now live, according to @0xd1namit (2026-05-11), who credits @BagCalls and his team for standout marketing and expects strong execution. The Funding Predicts Beta Competition (powered by Polymarket) launches with over 1.4m in funded accounts and 10K+ in cash prizes, giving traders 14 days to qualify at fundingpredicts.com.
By @0xd1namit
Why it matters
On 2026-05-10 23:35:53+00:00 Substack user Tpayneful liked the post "Founder’s Guide to Working with Filipino VAs".
Key details
- The liked post is authored by Spencer Burleigh; the notification originates from mg1.substack.com, shows "742" in the footer, and includes © 2026 Spencer Burleigh with a San Francisco mailing address (548 Market Street PMB 72296, CA 94104).
Brief
A Substack notification records that user Tpayneful liked Spencer Burleigh’s post "Founder’s Guide to Working with Filipino VAs" on 2026-05-10 23:35:53+00:00. The message was sent from mg1.substack.com, displays the number "742" in the footer, and carries a © 2026 Spencer Burleigh copyright plus a San Francisco mailing address.
By Tpayneful
Why it matters
SCOPE (Structured Decomposition and Conditional Skill Orchestration) maintains persistent semantic commitments via an evolving structured specification and conditionally invokes retrieval, reasoning, and repair skills to resolve the identified 'Conceptual Rift' in complex text-to-image intent realization.
Key details
- The paper introduces the human-annotated Gen-Arena benchmark with entity- and constraint-level specifications and the Entity-Gated Intent Pass Rate (EGIP). SCOPE achieves 0.60 EGIP on Gen-Arena and strong results on WISE-V (0.907) and MindBench (0.61).
Brief
SCOPE proposes a specification-guided framework that tracks semantic commitments across generation by decomposing intents into structured specifications and conditionally calling retrieval, reasoning, and repair skills to fix violations (the 'Conceptual Rift'). Evaluated on the new Gen-Arena benchmark with the EGIP metric, SCOPE yields 0.60 EGIP and also performs strongly on WISE-V (0.907) and MindBench (0.61). Summary based on the abstract (full text not provided).
Authors: Tianfei Ren, Zhipeng Yan, Yiming Zhao...
Why it matters
Proposes a Euclidean–Wasserstein-2 gradient-flow framework that jointly performs posterior sampling and prompt optimization in the latent space, aligning the generative prior and posterior with observed data.
Key details
- Combines the single flow with few-step latent text-to-image models to enable low-NFE inference without backpropagation through autoencoders, addressing the high-NFE and heavy backprop costs of prior LDM-based solvers (e.g., Rombach et al., 2022).
- Authors Alessio Spagnoletti, Tim Y. J. Wang, Marcelo Pereyra, and O. Deniz Akyildiz (arXiv:2605.07907v1, 2026-05-08) report state-of-the-art reconstructions on several canonical imaging inverse problems with significantly reduced computational cost.
Brief
The paper tackles inverse imaging with Vision–Language Latent Diffusion Models by introducing a unified Euclidean–Wasserstein-2 gradient-flow that jointly samples the posterior and optimizes prompts in latent space. By pairing this flow with few-step latent text-to-image models, the method achieves low-NFE inference and avoids backprop through autoencoders, yielding state-of-the-art results on canonical inverse problems with much lower compute; summary based on the abstract only.
Authors: Alessio Spagnoletti, Tim Y. J. Wang, Marcelo Pereyra...
Why it matters
On 2026-05-10 23:35:16+00:00 user Tpayneful liked the Substack post titled "Basic Capital's Big Idea: Bet Your Retirement on Interest Rates" (post by Spencer Burleigh).
Key details
- The notification email originates from mg1.substack.com; the item lists creation on 2026-03-11 12:44 and last updated 2026-04-04 00:00 and includes © 2026 Spencer Burleigh with a San Francisco mailing address.
Brief
Tpayneful liked a Substack post (published 2026-05-10 23:35:16+00:00) titled "Basic Capital's Big Idea: Bet Your Retirement on Interest Rates," attributed to Spencer Burleigh. The content is a platform notification (mg1.substack.com) showing metadata: created 2026-03-11 12:44, last updated 2026-04-04 00:00, and © 2026 Spencer Burleigh with a San Francisco address.
By Tpayneful
Why it matters
A purposive audit of 10 mechanistic-interpretability papers across four methodological strands (Lin & Liu, arXiv 2026-05-08) found no dedicated "identification-assumptions" section; papers commonly report validation metrics (faithfulness, completeness, monosemanticity, alignment, ablation effects) as causal support without stating the assumptions that would make those metrics identifying.
Key details
- A two-human-coder replication on n=30 reproduced the main direction: dedicated identification sections are absent and validation-metric substitution is common (exact counts are coding-rule sensitive). The authors propose a five-item disclosure norm: state if the claim is causal, name the identification strategy, enumerate assumptions, highlight at least one key assumption, and explain how conclusions change if assumptions fail.
Brief
Lin and Liu show that mechanistic interpretability work often uses causal vocabulary (circuits, mediators, causal abstraction) while omitting explicit identification assumptions. Via a purposive audit of 10 papers and a two-coder check on 30 items, they document widespread substitution of validation metrics for identification and offer a five-step disclosure norm to make causal claims explicit. Submitted to NeurIPS 2026 (Position Track).
Authors: Zezheng Lin, Fengming Liu
Why it matters
Method: Train a mean-pool Deep Set encoder on sets of size at most two to learn representations that generalize to arbitrary deployment set size N; then finetune the inference head on pre-aggregated embeddings so training compute/memory is essentially independent of N.
Key details
- Results & provenance: Authors Wehenkel, Kagan, Heinrich, and Pollard (arXiv 2026-05-08) evaluate on scalar, image, multi-view 3D, molecular, and high-dimensional conditional-generation benchmarks with N in the thousands, matching or outperforming standard baselines at a fraction of the compute.
Brief
It Just Takes Two introduces a simple, theoretically grounded strategy for amortized neural posterior estimation that decouples representation learning from posterior modeling. The authors train a mean-pool Deep Set on sets of size ≤2 to produce an encoder that generalizes to arbitrary set sizes, then finetune an inference head on aggregated embeddings; this makes training cost essentially independent of deployment set size N. Across diverse benchmarks (scalar, image, multi-view 3D, molecular, high-dimensional generation) with N in the thousands, the method matches or outperforms baselines while using much less compute.
Authors: Antoine Wehenkel, Michael Kagan, Lukas Heinrich...
Why it matters
Andrew Ng announced on 2026-05-11 that Coursera and Udemy have merged into a single company and he will serve as Chairman, working alongside Greg Hart and the combined leadership team.
Key details
- Ng asserts the merger combines broader learning content, trusted instructors, and more engaging experiences to make learning more personalized, applied, and accessible at scale.
- He frames the move as critical because AI is changing the nature of work and increasing demand for continuous, job-relevant skill-building worldwide.
Brief
Andrew Ng announced on May 11, 2026 that Coursera and Udemy have joined as one company and named him Chairman alongside Greg Hart. He argues the combined platform will unite content, instructors, and experiences to deliver more personalized, applied, and accessible learning at scale to meet rising AI-driven demand for job-relevant skills.
By @AndrewYNg
Why it matters
Royal Pop is a Pump.fun Solana token tied to the Swatch x Audemars Piguet Royal Oak collab dropping May 16; it's ~1 day old with an estimated $130K–$240K market cap (ATH $247K), ~$380K 24h volume, ~41K in liquidity and ~450 holders.
Key details
- The watch (bioceramic Royal Oak) retails ~$300–$500, in-store only, with global queues forming; watch press frames this as 'MoonSwatch 2.0' and event-driven tokens around confirmed luxury catalysts have historically done 10–50x in days when timing/framing align.
- Significant downside risk: author notes >95% of Solana tokens go to zero, multiple copycats (each < ~$20K) and meme risk ('Royal Poop') exist; Royal Pop could 3–4x or halve in hours, 10–20x into May 16 is plausible, but post–May 20 odds fall unless a real community (like Labubu) forms—99% of top-50 holders bought their own supply.
Brief
Royal Pop is a Pump.fun Solana meme token built around the Swatch × Audemars Piguet Royal Oak release on May 16; it's trading with ~$130K–$240K market cap, ~$380K 24h volume and ~450 holders. The author argues momentum and MoonSwatch framing could drive a 10–20x run into launch, but >95% of Solana tokens fail and post–May 20 value depends on forming a real community.
By @ianmiles
Why it matters
Visa Business Card Companion Fare Offer in Spencer Burleigh’s Alaska Airlines (Atmos Rewards) account expires on August 8, 2026 (approximately three months from the 2026-05-10 notice); the account was last updated 101 days ago.
Key details
- Alert published by AwardWallet on 2026-05-10 advises logging into AwardWallet to verify the expiration and use auto-login links; message includes unsubscribe and account-restore links and AwardWallet contact/address details.
Brief
The Visa Business Card Companion Fare Offer in Spencer Burleigh’s Alaska Airlines (Atmos Rewards) account is set to expire on August 8, 2026 (about three months after the 2026-05-10 AwardWallet notice). AwardWallet reports the account was last updated 101 days ago and urges logging into AwardWallet to verify the coupon using provided auto-login and account-restore links.
By AwardWallet
Why it matters
TESS has an estimated >300,000 oscillating red giants with mostly 1–2 month observations; the authors develop a deep-learning method to infer global seismic parameters from such short-duration data.
Key details
- On one-month Kepler and K2 samples the ML algorithm recovers Δν and ν_max accurately for ≈50% of targets; for one-sector TESS data reliable Δν is recovered for only ≈23% of stars.
- From K2 the method yields reliable dipolar period spacings (ΔΠ1) for ≈200 young red giants, reproducing the well-known Δν–ΔΠ1 degenerate sequence seen in Kepler red giants.
Brief
The authors apply deep learning to infer global asteroseismic parameters (Δν, νmax, and for K2 also ΔΠ1) from short, one-month lightcurves to enable scalable analysis of TESS/K2 red-giant samples. Their model recovers Δν and νmax for ~50% of one-month Kepler/K2 cases but only ~23% for single-sector TESS; it produces ~200 reliable ΔΠ1 measures that match the Kepler Δν–ΔΠ1 sequence. (Summary based on the abstract; full text was not provided.)
Authors: Nipun Ghanghas, Siddharth Dhanpal, Shravan Hanasoge...
Why it matters
On 2026-05-10T23:35:29+00:00 Substack user "Tpayneful" liked the post titled "Dancing with Missiles".
Key details
- The notification originates from mg1.substack.com, references Spencer Burleigh (© 2026), and is an engagement email containing layout and image assets rather than the article text.
Brief
A Substack notification dated 2026-05-10 23:35:29+00:00 showing that user 'Tpayneful' liked the post 'Dancing with Missiles'. The message, delivered via mg1.substack.com and credited to Spencer Burleigh (© 2026), is an engagement/like email containing layout and image assets rather than substantive article content.
By Tpayneful
Why it matters
Tool identity is linearly readable and steerable: adding the mean-difference between two tools' average internal activations flips the model's chosen tool with 77–100% accuracy on name-only single-turn prompts (93–100% for models ≥4B), and the autoregressive JSON arguments follow the new tool's schema.
Key details
- The causal effect concentrates on the output row for the target tool and a small set of mid/late-layer attention heads: injecting a unit vector along that output-row reaches 93–100% success; activation patching localises responsibility to those heads; a within-topic probe across 14 airline tools achieves 61–89% top-1 on five 4B–14B models.
- Pretraining encodes tool identity before generation: cosine readout from base models recovers 69–82% tool identity while base generation is only 2–10%; model suite tested includes 12 instruction-tuned models (Gemma 3, Qwen 3, Qwen 2.5, Llama 3.1) from 270M to 27B. Also, on Gemma 3 12B/27B, queries with smallest top-1 vs top-2 activation gaps produce 14–21× more wrong calls.
Brief
The paper shows that language models (270M–27B, including Gemma 3, Qwen, Llama 3.1) encode tool selection as a linearly readable vector: adding per-tool mean-difference vectors reliably flips name-only single-turn tool choices and causes downstream JSON arguments to match the new schema. Causal attribution concentrates on one output-row and a few mid/late attention heads; base-model representations already carry tool identity (69–82% recoverable), while instruction tuning wires it to generation. Measurements are for single-turn fixed-menu settings; multi-turn transfer is noted as more fragile.
Authors: Zekun Wu, Ze Wang, Seonglae Cho...
Why it matters
123D is an open-source framework that unifies multi-modal autonomous-driving data under a single API by storing each sensor/modal modality as an independent timestamped event stream (no prescribed rate), enabling synchronous or asynchronous access across heterogeneous datasets; it consolidates eight real-world datasets totaling 3,300 hours and 90,000 kilometers plus a configurable synthetic dataset.
Key details
- The authors use 123D to systematically compare annotation statistics and evaluate pose/calibration accuracy across datasets, and showcase two applications enabled by the framework: cross-dataset 3D object-detection transfer and reinforcement-learning for planning; code and docs are at https://github.com/kesai-labs/py123d.
Brief
123D unifies multi-modal driving datasets by representing each sensor or annotation as a timestamped event stream, allowing flexible synchronization across heterogeneous formats. The authors merge eight real-world datasets (3,300 hours, 90,000 km) and a synthetic generator, perform systematic analyses of annotations and pose/calibration, and demonstrate cross-dataset 3D detection transfer and RL planning; the framework and tools are released open-source.
Authors: Daniel Dauner, Valentin Charraut, Bastian Berle...
Why it matters
Karpathy (created 2026-03-11) recommends ending LLM prompts with the exact phrase 'structure your response as HTML' to view generated output in a browser and reports similar success asking for slideshow-style output.
Key details
- He asserts audio is humans' preferred input and vision is AIs' preferred output, noting ~one-third of the brain is dedicated to vision, and predicts a progression: raw text → markdown → HTML → interactive neural videos/simulations, ultimately from diffusion neural nets.
- He urges better multimodal inputs (pointing/gesturing on-screen) before moving to Neuralink‑style BCIs, calling the current phase an ongoing 'input/output mind‑meld' with substantial work remaining.
Brief
Karpathy recommends appending 'structure your response as HTML' to LLM prompts to render outputs in a browser, argues audio will be humans' preferred input while vision becomes AI's preferred output, and outlines a progression from text→markdown→HTML→interactive neural videos (eventually diffusion‑generated), urging improved multimodal inputs before BCIs.
By @karpathy
Why it matters
Alex Luposasca (guest on Latent.Space, episode published 2026-05-11) reports GPT‑5.x derived new results in scattering amplitudes, including a simplification for single‑minus gluon tree amplitudes (discussed ~14:38–31:26) and a reconstructed proof from scratch (~38:07) to verify validity.
Key details
- GPT‑5 helped solve a year‑long physics puzzle (~20:56–23:02); GPT Pro then generalized those techniques to graviton amplitudes (~42:27–53:57). Luposasca also credits GPT‑5 with solving a black‑hole perturbation problem (~1:12:46).
- Luposasca characterizes AI as a 'scout' and collaborator that accelerates theoretical discovery but warns of 'AI slop' risks to publishing quality and says the bottleneck is shifting to human curation, taste, and writing papers (~53:57–1:30:19).
Brief
Alex Luposasca, on Latent.Space (published 2026-05-11), describes using GPT‑5.x to derive new scattering‑amplitude results (single‑minus gluon trees), solve a year‑long puzzle, generalize methods to graviton amplitudes, and crack a black‑hole perturbation problem. He portrays AI as a productive research 'scout' while warning of hallucination risks and a new bottleneck in paper writing and curation.
By @ALupsasca
Why it matters
Garry's List published the Action Voter Guide on 2026-05-11 (authors: Garry Tan, Shaudi Fulp, Forrest Liu) compiling endorsements for the California primary on June 2, 2026.
Key details
- The guide synthesizes recommendations from local housing groups, labor unions, and civic reformers, is labeled 'transparent, sourced, searchable' at garrysguide.org/elections, and invites additions via hello@garryslist.org or X.
Brief
The Garry's List Action Voter Guide, published May 11, 2026 by Garry Tan, Shaudi Fulp, and Forrest Liu, aggregates endorsements from local housing groups, labor unions, and civic reform organizations ahead of the California June 2 primary. The resource is presented as transparent, sourced, and searchable at garrysguide.org/elections and accepts suggested additions via email or X.
By Garry Tan, Shaudi Fulp
Why it matters
Darius Foroux (Substack, published May 11, 2026) says that after spending his adult life building passive income he still experiences anxiety — the worry didn’t disappear, it simply shifted to new targets.
Key details
- He invokes psychiatrist Gordon Livingston and Livingston’s book And Never Stop Dancing to argue that life’s complexity produces persistent unwanted emotions and that eliminating the ‘burden of striving’ raises the question, “What relevance do we retain?”
- Foroux proposes a behavioral remedy: treat passive income as a practical foundation but prioritize daily, slightly uncomfortable habits (examples: write when you don’t feel like it, work out, do taxes, mow the lawn) to stay sharp and derive satisfaction; he recounts mowing his overgrown grass and feeling genuinely good afterward.
- His conclusion: financial freedom buys time and options but is the wrong sole organizing goal — continued striving, contribution, and small daily challenges preserve purpose and reduce anxiety better than passive income alone.
Brief
Darius Foroux, writing on Substack (published May 11, 2026), reflects on decades spent building passive income and concludes that money alone did not remove anxiety — it only redirected it. Drawing on Gordon Livingston’s And Never Stop Dancing, Foroux frames the problem as a human response to life’s complexity and warns that eliminating striving threatens one’s sense of relevance. He recommends treating passive income as a foundation rather than an end and cultivating daily, slightly uncomfortable habits to stay engaged: examples include writing when unmotivated, exercising, doing taxes, and his recent anecdote of mowing overgrown grass and feeling rewarded. Foroux’s practical prescription is to use financial freedom to buy time, then invest that time in continued growth, contribution, and routines that preserve purpose and mental sharpness.
By Darius Foroux
Why it matters
On 2026-05-10 23:36:30+00:00 Substack user 'Tpayneful' liked the post 'Founder’s Guide to Hiring an Operations Team in the Philippines' (post hosted on mg1.substack.com).
Key details
- Notification metadata shows the item was created 2026-03-11-12-44 and last updated 2026-04-04; the footer credits Spencer Burleigh © 2026 with mailing address 548 Market Street PMB 72296, San Francisco, CA 94104.
- The Substack notification UI displays a numeric '261' indicator and includes standard controls (view profile, mute post) with images and assets served from substackcdn.com.
Brief
Notification records that Substack user Tpayneful liked 'Founder’s Guide to Hiring an Operations Team in the Philippines' on 2026-05-10 23:36:30+00:00. The email-style notice includes creation timestamp 2026-03-11-12-44, last updated 2026-04-04, author credit Spencer Burleigh © 2026, San Francisco mailing address, and UI elements showing '261' and profile/mute links.
By Tpayneful
Why it matters
Proposes a retrieval-guided diffusion noise optimization method that mixes retrieved noise with random noise via a reward-guided mask to better initialize diffusion sampling, and claims this enables generation that satisfies highly-constrained spatiotemporal goals (e.g., severe spatial obstacles or specified numbers of walking steps).
Key details
- Adds relational task parsing (using an LLM) to identify the hardest constraints and select retrieval references; the framework is training-free and was released on arXiv 2026-05-08 (authors: Hanchao Liu et al.), and the paper is accepted to CVPR 2026.
Brief
The paper addresses zero-shot human motion generation under very challenging spatiotemporal constraints by augmenting training-free diffusion noise optimization with retrieval guidance. It parses task constraints into groups (relational task parsing, powered by an LLM), retrieves reference motions for the hardest constraints, and forms a reward-guided mask to blend retrieved and random noise for improved diffusion initialization. The authors report this approach successfully handles tasks that prior methods struggle with, enabling more reliable constrained motion synthesis; accepted to CVPR 2026 (arXiv 2026-05-08).
Authors: Hanchao Liu, Fang-Lue Zhang, Shining Zhang...
Why it matters
CMR-EXTR converts free-text cardiac magnetic resonance (CMR) reports into auditable structured data and provides per-field confidence; evaluated results report 99.65% variable-level accuracy (ArXiv: 2026-05-08; authors include Yi Yu and Parker Martin).
Key details
- The system uses a teacher–student distillation pipeline to enable fully offline inference with limited manual annotation and an uncertainty scheme combining distribution plausibility, sampling stability, and cross-field consistency to triage human review.
- Authors claim this is the first CMR-specific extraction system with integrated confidence estimation; code is available at https://github.com/yuyi1005/CMR-EXTR and the work was accepted to ISBI 2026.
Brief
CMR-EXTR converts free-text cardiac magnetic resonance (CMR) reports into auditable structured data with per-field confidence. The method employs a teacher–student distillation pipeline for fully offline inference and reduced annotation effort, plus an uncertainty model that blends distribution plausibility, sampling stability, and cross-field consistency to prioritize human review. It achieves 99.65% variable-level accuracy; code on GitHub; accepted to ISBI 2026.
Authors: Yi Yu, Parker Martin, Zhenyu Bu...
Why it matters
Lee, Mehrotra, and Zampetakis (arXiv 2026-05-08) prove that every class of sets with Gaussian surface area ≤ Γ admits non-negative degree-k polynomials that ε-approximate its indicator function in L1 under the standard Gaussian, with k = ~O(Γ^2/ε^2).
Key details
- The approximants have range contained in [0,∞) (a stronger pointwise guarantee than ordinary L1-approximation but weaker than sandwiching polynomials), match the best known Gaussian L1-approximation degree up to constant factors, and are motivated by applications to smoothed learning from positive-only examples.
Brief
The note proves that finite Gaussian surface area Γ implies existence of non-negative degree-k polynomials that ε-approximate indicator functions in L1 under the standard Gaussian, with k = ~O(Γ^2/ε^2). This adds a pointwise non-negativity guarantee (range [0,∞)), sits between plain L1-approximation and sandwiching polynomials, matches prior degree bounds up to constants, and targets smoothed positive-only learning. Only the abstract was available for this summary.
Authors: Jane H. Lee, Anay Mehrotra, Manolis Zampetakis
Why it matters
The authors derive two Q-value–style extensions of the Bellman equation for exponential-utility optimization in discounted MDPs and show the associated operators are contractions in the L_infty and sup-log/Thompson metrics; they characterize fixed points and prove the induced greedy stationary policy is optimal among stationary policies.
Key details
- They propose two model-free algorithms: a two-timescale Q-learning–style method with almost-sure convergence and finite-time convergence rates obtained via timescale separation, and a one-timescale algorithm driven by a sublinear power-law operator that lacks a global contraction but is shown to converge using local Lipschitzness, monotonicity, homogeneity, and Dini-derivative arguments (scalar finite-time analysis only).
- Preprint: Gugan Thoppe, L. A. Prashanth, Ankur Naskar, Sanjay Bhat; arXiv:2605.08053v1 (cs.LG), published 2026-05-08; builds on Bellman-type exponential-utility work (e.g., Porteus 1975) to provide a foundation for value-based RL under fixed risk-aversion.
Brief
Reinforcement-learning for exponential-utility optimization in discounted MDPs: the paper derives two Q-value–style Bellman extensions whose operators are contractions in L_infty and sup-log/Thompson metrics, proves fixed-point structure and optimality of the induced greedy stationary policy among stationary policies, and presents two model-free algorithms — a two-timescale Q-learning with a.s. convergence and finite-time rates, and a one-timescale power-law method whose convergence is established via delicate local arguments. Full text on arXiv (abstract used).
Authors: Gugan Thoppe, L. A. Prashanth, Ankur Naskar...
Why it matters
Chevy Chase told a female interviewer, “No sh*t?! You’re not bright enough,” then added, “I know you’re not gonna put that on the air… and I hope not,” in a clip from Marina Zenovich’s film I’m Chevy Chase and You’re Not (video credited to SkyTV/CNN) that went viral on X on May 11, 2026.
Key details
- @HustleBitch_ posted the clip (May 11, 2026) framing it as “CHEVY CHASE COMPLETELY INSULTS INTERVIEWER,” asked “Is he misunderstood… or just an asshole?”, and noted commenters say the exchange exemplifies Chase’s long-standing contentious reputation.
Brief
Chevy Chase insulted a female interviewer in a clip from Marina Zenovich’s I’m Chevy Chase and You’re Not, saying “No sh*t?! You’re not bright enough” and “I know you’re not gonna put that on the air… and I hope not.” A viral X post by @HustleBitch_ (May 11, 2026) highlighted the exchange and linked it to Chase’s contentious reputation.
By @HustleBitch_
Why it matters
Real Brokerage agreed to acquire RE/MAX Holdings for approximately $880 million in a deal expected to close in H2 2026; RE/MAX shareholders can elect $13.80 cash or 5.152 shares of the new Real RE/MAX Group, with Real shareholders retaining ~59% ownership and the combined company reporting pro forma 2025 revenue of $2.3 billion and projected $30 million in annual cost savings by end of 2027.
Key details
- GEM's Proptech Index (25 stocks) had a combined market cap of $234.758 billion as of 5/8/2026, up 0.51% week-over-week.
- The first batch of Q1 earnings summaries published this week cover Zillow, CoStar, ProCore and Compass (full set of ten summaries to be released to Crystal subscribers once complete); exclusive decks and summaries are available to VCs/angels via application to community@geekestate.com.
- GEM updates and events: Weekly Radar #400 (published 11 May 2026) marks 400 radars and 350 transmissions; upcoming member events include Fundraising office hours with Ryan Coon on 5/14 and an Innovators Roundtable Dinner during the Housing Innovation Alliance summit in Charlotte on 5/19; active fundraising deals listed include America Housing Corp ($50M Series A), LotRoll ($3M Seed), Zavvie (10% mezzanine), and Quiet Cove ($750k Seed).
Brief
GEM's Weekly Radar #400 (published 11 May 2026) combines market intel, member news and sector analysis across proptech and real estate. Key macro data: the GEM Proptech Index (25 stocks) reached a $234.758B combined market cap as of 5/8/2026 (+0.51% w/w). Earnings coverage for Q1 began with summaries of Zillow, CoStar, ProCore and Compass (ten-company series planned for subscribers). Strategically significant M&A: Real Brokerage will acquire RE/MAX for ~ $880M to form Real RE/MAX Group (RE/MAX holders may take $13.80 cash or 5.152 shares), creating a combined pro forma 2025 revenue base of $2.3B, >180,000 agents across 120+ countries, and expected $30M annual run-rate savings by end of 2027. The newsletter also lists active fundraising opportunities (e.g., America Housing Corp $50M Series A) and upcoming GEM member events on 5/14 and 5/19.
By Crystal, Powered by GEM
Why it matters
On May 11, 2026, founder Jason "The Memelord" Levin announced that the Memelord app has "AGI for memes" and claimed the capability is available inside the app
Key details
- The newsletter corrected a previously wrong download/link, directs readers to the corrected app link and Memelord.com, offers a free trial, and includes promo code MYTHOS for 50% off
Brief
Memelord’s May 11, 2026 newsletter from founder Jason “The Memelord” Levin claims the Memelord app has in‑app 'AGI for memes' and corrects a prior wrong download/link. The note links to the app, promotes a free trial at Memelord.com, and offers promo code MYTHOS for 50% off signups.
By Memelord Magazine
Why it matters
KKR committed $300 million to FS KKR Capital Corp. (FSK): $150 million in convertible preferred (5% cash / 7% PIK) convertible at $18.83 (FSK’s quarter-end NAV) and a $150 million tender offer for common shares at $11 per share (FSK stock closed the prior week at $10.84).
Key details
- Public business-development companies (BDCs) are trading below NAV; Apollo is in talks to sell MidCap Financial Investment Corp. (MFIC), whose stock trades ~85% of NAV, likely in a share-for-share deal with another BDC rather than cash at full NAV.
- Private credit lending volume fell 14% in Q1 2026 while US banks’ lending rose 12.7% (the fastest growth since 2022), amid higher funding costs for private credit and regulatory easing that benefits banks, per OCC commentary.
- Virginia’s redistricting referendum received 51.7% 'Yes' (48.3% 'No') in April but the Virginia Supreme Court later voided the referendum process; Kalshi’s prediction markets ($5.7M and ~$10M of bets) resolved inconsistently because an accelerated-resolution clause used four of eight designated media (NYT, AP, DDHQ, CNN, Fox, NBC, CBS, ABC) to settle one contract in April.
Brief
KKR’s $300 million intervention in FS KKR Capital Corp. (FSK) typifies how private-credit sponsors are responding to persistent BDC discounts to NAV. KKR will put in $150 million of convertible preferred—paying 5% cash or 7% in kind—with a conversion price equal to FSK’s quarter-end NAV ($18.83), and simultaneously tender for $150 million of common at $11 per share (FSK traded at $10.84). The package mixes an NAV-linked show of confidence with an economically accretive purchase of discounted shares. The move comes as many public BDCs trade below reported NAVs and dealmaking (e.g., Apollo’s talks to sell MFIC) increasingly uses stock-for-stock structures because buyers won’t pay full cash NAV.
Broader market context: private credit originations fell ~14% in Q1 2026 while bank lending jumped 12.7%, reflecting higher funding costs for private lenders and deregulatory shifts that let banks compete on leveraged lending. Separately, a legal-technicality produced conflicting outcomes in prediction markets: Virginia’s redistricting referendum drew 51.7% 'Yes' in April but was later voided by the Virginia Supreme Court; Kalshi’s contracts resolved differently because an accelerated oracle clause relied on four of eight designated media to declare a winner. And in tech labor markets, OpenAI’s expanded tender cap (to $30M per person) let ~75 employees hit the cap, illustrating how private liquidity is reshaping compensation and retention dynamics in AI firms.
By Matt Levine
Why it matters
Heathrow Terminal 3 hosts at least ten airside lounges (including oneworld lounges from American Airlines, British Airways, Cathay Pacific and Qantas, plus Emirates and Virgin Atlantic Clubhouse), making it a candidate for 'peak lounge' density (Oliver Ranson, May 11, 2026).
Key details
- UK’s four largest airports (Heathrow, Gatwick, Stansted, Manchester) generated £673 million (~$915M) from car parking in 2023 — roughly £1.8M (~$2.5M) per day — and Heathrow added 900 parking spaces in 2024; non-aeronautical revenue rose from 36.3% of airport revenue in Q1‑23 to 37.2% in Q1‑24.
- Ranson proposes monetising the departures environment by charging for access to retail and hospitality beyond basic gates/amenities (while keeping basic seating, toilets and gate access free), noting risks such as tenants demanding lower rents, passengers arriving later (increasing pressure on check‑in/security), and operational delays.
- The article classifies lounge models (airport free departures areas; pay‑to‑access airport lounges; third‑party operators such as Collinson with >80 venues; airline‑branded lounges often outsourced; and rare special terminals like Lufthansa First at Frankfurt) and flags special terminals as high‑opportunity for monetisation.
Brief
Airport lounges are nearing saturation in some hubs — Heathrow Terminal 3 alone offers at least ten airside lounges (oneworld’s American, British Airways, Cathay Pacific and Qantas among them, plus Emirates and Virgin’s Clubhouse) — prompting Oliver Ranson (May 11, 2026) to ask whether we’ve hit “peak lounge.” He outlines five lounge categories (free airport departure areas; paid airport lounges; third‑party operators such as Collinson with 80+ venues; airline‑branded lounges; and rare premium terminals like Lufthansa First) and focuses on monetising the ordinary departures hall. Drawing parallels to car‑park pricing (UK top airports earned £673M from parking in 2023; Heathrow added 900 spaces in 2024) and rising non‑aeronautical share (36.3% → 37.2% from Q1‑23 to Q1‑24), Ranson suggests gating retail/hospitality behind paid access while keeping essential gate access free, and warns of tenant pushback, later passenger arrivals, and pressure on check‑in/security.
By Oliver Ranson from Airline Revenue Economics
Why it matters
Astral Codex Ten published Open Thread 433 on May 11, 2026 (Scott Alexander) as the weekly visible open thread with links to the ACX subreddit, Discord, bulletin board, and in-person meetups.
Key details
- A pending US Congress agriculture bill would preempt existing state animal‑welfare laws — for example revoking California’s ban on keeping pigs in crates too small to turn around — and animal‑welfare groups call it “the most important legislative threat to farmed animal welfare in US history.”
- Hampshire College professor Ethan Ludwin‑Peery warned the campus is shutting down and issued a rescue plea estimating $30–$60 million to buy the campus plus about $40 million to operate it for the first few years; he invited interested parties to email ethanludwinpeery@gmail.com.
Brief
Open Thread 433 (Astral Codex Ten, published May 11, 2026) is the weekly community open thread by Scott Alexander that aggregates reader items and links to ACX social channels. Two top highlights: (1) a currently considered US agriculture bill that would federally preempt all state animal‑welfare protections — the post cites California’s sow‑crate prohibition as an example — and animal‑welfare organizations are framing the measure as the single biggest legislative threat to farmed‑animal protections in US history, with calls to contact Senators. (2) Ethan Ludwin‑Peery, a former book‑review contest winner and current Hampshire College professor, says administrative financial errors have left the college facing closure and requests $30–60M to acquire the campus plus ~$40M to fund operations during a transitional period, providing his email for inquiries.
By Astral Codex Ten
Why it matters
The inaugural Cerebral Valley Voice Summit (May 2026) gathered 200+ founders, investors, and operators and hosted 13 talks from leaders including Bret Taylor (Sierra), Justin Uberti (OpenAI), Anastasis Germanidis (Runway) and founders from Abridge, Wispr Flow, Deepgram, Cartesia and more.
Key details
- Sierra’s Bret Taylor spoke two days after securing a fresh $950M financing, reporting >$165M in revenue and adoption by 40% of the Fortune 50; he called voice AI ‘early innings’ compared to the internet pre‑broadband.
- OpenAI’s Justin Uberti and the company’s realtime team released new realtime voice models that can perform reasoning during a conversation; panelists debated tradeoffs between highly human‑like speech and accuracy/clarity for task‑oriented agents.
- Healthcare and enterprise traction were highlighted: Assort Health reports ~150 million patient interactions across 5,000 providers via voice agents; startups noted low‑latency TTS (Cartesia), the value of shaving milliseconds for inference (LiveKit), and the rise of multimodal/world models (Runway, MiniMax).
Brief
The Cerebral Valley Voice Summit convened 200+ builders and investors in May 2026 for 13 recorded panels on voice AI, featuring Bret Taylor (Sierra), Justin Uberti (OpenAI), Anastasis Germanidis (Runway) and founders from Abridge, Wispr Flow, Deepgram and Cartesia. Taylor announced Sierra’s recent $950M raise, >$165M revenue and deployments at 40% of the Fortune 50, while OpenAI released realtime voice models capable of reasoning mid‑conversation. Panels debated whether agents should sound fully human versus prioritizing accurate, task‑oriented behavior; Deepgram’s Scott Stephenson said voice models hadn’t yet passed his five‑minute “voice Turing Test” but predicted that context memory advances could do so by year‑end. Speakers also emphasized enterprise wins in healthcare (Assort Health: ~150M interactions across 5,000 providers), low‑latency TTS, the importance of latency reductions for inference, and a near‑term shift toward multimodal/world models that fuse voice with video and motion.
By Newcomer
Why it matters
CA-SQL achieves a state-of-the-art 51.72% execution accuracy on the "challenging" tier of the Bird-Bench (BIRD) development set using only GPT-4o-mini (reported 2026-05-08).
Key details
- The pipeline uses complexity-aware, difficulty-scaled exploration, evolutionary-search-inspired prompt seeding, and a novel voting selector; overall results on BIRD dev are 61.06% execution accuracy and 68.77% Soft F1.
Brief
CA-SQL is a complexity-aware Text-to-SQL inference pipeline that scales exploration breadth by estimated task difficulty, employs evolutionary-search-inspired prompt seeding to elicit diverse candidates, and uses a novel voting method to pick final queries. On the Bird-Bench development set it reports 51.72% on the "challenging" tier (SOTA among in-context approaches with GPT-4o-mini), plus 61.06% execution accuracy and 68.77% Soft F1.
Authors: James Petullo, Nianwen Xue
Why it matters
In December 2025 President Donald Trump pardoned former Honduran president Juan Orlando Hernández, who had been serving a 45-year prison sentence for drug trafficking.
Key details
- On April 30, 2026 Canal Red published leaked audio recordings of unclear origin and unverified authenticity in which a voice purported to be Hernández alleges Israel financed and brokered his release while Trump allies would smooth his political return in exchange for Honduras expanding special economic zones, hosting a new US military base, and passing laws favorable to American and Israeli corporate interests.
- The leaks claim a broader influence operation — including a disinformation campaign against Mexico and Colombia — allegedly funded in part by Argentina’s president Javier Milei; Hernández denies the tapes, and neither Trump nor Israel has publicly responded.
- Mexican president Claudia Sheinbaum acknowledged the report but minimized its likely effect, while Colombia’s Gustavo Petro cited the recordings as evidence of efforts to undermine progressive governments; independent verification of the recordings remains absent, underscoring institutional fragility in parts of Central America.
Brief
Honduras-gate centers on a set of leaked audio files published on April 30, 2026 by Canal Red that purport to record Juan Orlando Hernández describing a deal in which Israel helped finance and broker his release and US-aligned actors would facilitate his political comeback after Donald Trump’s December 2025 pardon removed a 45-year drug-trafficking sentence. The tapes allege concrete quid pro quos — expanded special economic zones, a new American military base and laws favoring US and Israeli corporate interests — and describe a wider campaign including disinformation against Mexico and Colombia allegedly funded in part by Argentina’s Javier Milei. Hernández denies the recordings, Trump and Israel have not replied, and no independent verification of provenance or authenticity has been published, leaving the episode as both a potential example of interstate influence operations and a marker of Central America’s fragile democratic institutions.
By The Economist
Why it matters
Collected paired listened and imagined MEG from trained musicians listening to rhythmic, melodic, and spoken stimuli and trained six linear and neural mapping models to predict listened responses from imagined MEG.
Key details
- Trained a contrastive word decoder solely on listened MEG using four embedding strategies (including semantic, acoustic, phonetic); applying the mapped imagined→listened responses from held-out subjects produced word decoding significantly above chance by rank-based analysis, and performance improved with more training data.
Brief
Imagined speech decoding: the authors recorded paired listened and imagined MEG from trained musicians and built a three-stage pipeline that maps imagined MEG to listened MEG (six mapping models), decodes words with a contrastive listened-only decoder (four embedding strategies), and applies the decoder to mapped imagined data from held-out subjects. Proof-of-concept results show significant above-chance decoding and scalability with more training data; only the abstract was available for this summary.
Authors: Maryam Maghsoudi, Shihab Shamma
Why it matters
GRAPHLCP is a structure-aware, proximity-based localized conformal prediction framework for GNNs introduced by Peyman Baghershahi, Fangxin Wang, Debmalya Mandal, and Sourav Medya (arXiv 2026-05-08).
Key details
- The method adds a feature-aware densification step and a Personalized PageRank (PPR)-based kernel for topology-dependent anchor sampling and calibration weighting, explicitly modeling local and long-range graph dependencies.
- GRAPHLCP provably guarantees marginal coverage with finite samples and, according to experiments on multiple regression and classification datasets, attains improved test conditional coverage and more efficient prediction sets versus embedding-only localization (paper: 20 pages, 9 figures, 8 tables).
Brief
GRAPHLCP tackles conformal prediction for graph neural networks by addressing failures of embedding-space localization on graphs. It combines feature-aware densification to reduce locality bias in sparse graphs with a Personalized PageRank kernel to capture structural proximity for anchor sampling and calibration weighting. The approach yields finite-sample marginal coverage and empirically better test conditional coverage and smaller prediction sets on several regression and classification benchmarks compared to prior embedding-only localization methods.
Authors: Peyman Baghershahi, Fangxin Wang, Debmalya Mandal...
Why it matters
Expanding context windows across 7 LLMs and 4 games over 500 rounds degraded cooperation in 18 of 28 model–game settings (Liu et al., 2026).
Key details
- Mechanism analyses on 378,000 reasoning traces attribute collapse to eroded forward-looking intent (not paranoia); a LoRA adapter fine-tuned on forward-looking traces mitigates the decay and transfers zero-shot, memory sanitization (replacing history with synthetic cooperative records) restores cooperation, and ablating explicit Chain-of-Thought often reduces the collapse.
Brief
The paper 'The Memory Curse' (Liu et al., 2026) shows that expanding LLMs' context windows often erodes cooperation in multi-agent social dilemmas: across 7 models and 4 games over 500 rounds cooperation fell in 18 of 28 model–game settings. Analyses of 378,000 reasoning traces implicate loss of forward-looking intent; targeted LoRA fine-tuning, memory sanitization, and CoT ablations partly restore cooperation. (Based on the abstract.)
Authors: Jiayuan Liu, Tianqin Li, Shiyi Du...
Why it matters
Energy secretary Chris Wright said the Trump administration would back suspending the federal petrol tax (currently 18¢ per gallon) as average US pump prices sit about $4.34/gal; any suspension would require Congressional approval and Wright withdrew a March prediction that prices would fall below $3 by summer.
Key details
- President Donald Trump publicly attacked two of his appointees on the Supreme Court, Neil Gorsuch and Amy Coney Barrett, saying “it’s really OK for them to be loyal” after they voted to strike down some of his tariffs.
- Virginia’s recently drawn Democratic-leaning congressional map was struck down by the state supreme court; Democrats held a testy call (reported by the New York Times) weighing responses including imposing a judicial age limit or launching a fresh challenge to the state’s independent-redistricting law.
- Two Americans from a Dutch cruise-ship hantavirus outbreak tested positive or have mild symptoms after the ship disembarked in Spain’s Canary Islands; they are being airlifted to a specialised centre at the University of Nebraska in Omaha.
Brief
The US in Brief rounds up fast-moving political and domestic developments: Energy secretary Chris Wright said the administration would support suspending the federal petrol tax (18¢/gal) to ease pump pain—average prices are about $4.34/gal—but Congress must act and Wright retracted a March expectation of sub-$3 summer fuel. On politics, President Trump publicly blasted two of his Supreme Court appointees, Neil Gorsuch and Amy Coney Barrett, for recent tariff rulings, urging loyalty. In Virginia, the state supreme court struck down a Democratic-favouring congressional map; party leaders are considering measures from imposing judicial age limits to fresh legal challenges to independent redistricting. Health officials are managing a hantavirus cluster from a Dutch cruise-ship visit to the Canary Islands, flying affected Americans to the University of Nebraska’s specialised centre in Omaha. Also, Abe Foxman, longtime head of the ADL, died at 86.
By The Economist
Why it matters
NoiseGate (Wen Huang et al., arXiv:2605.07794v1, published 2026-05-08) replaces the common single shared timestep t in Mixture-of-Transformers (MoT) world-action models with learnable per-latent timestep schedules, treating each predicted latent frame's noise level as an information-gating policy.
Key details
- The method combines independent per-latent timestep sampling during backbone training, a lightweight Gating Policy Network that emits per-latent time increments during denoising, and task-reward optimization to train schedules without hand-crafted shape priors.
- Built on a joint video–action MoT backbone, NoiseGate yields consistent gains on diverse RoboTwin random-scene manipulation tasks (reported in the paper's abstract).
Brief
NoiseGate reframes per-latent timestep selection in joint video–action world-action models as a learnable information-gating policy: by adjusting each predicted latent frame's noise level, a Gating Policy Network controls its Key/Value reliability for action generation. The approach (independent per-latent timestep sampling plus task‑reward optimization) improves RoboTwin random-scene manipulation performance. Summary based on the paper's abstract; full text not available here.
Authors: Wen Huang, Haoran Sun, Yongjian Guo...
Why it matters
Matthew Yglesias (Slow Boring, May 11, 2026) flags a hantavirus outbreak linked to the cruise ship MV Hondius with five confirmed cases and three deaths reported in coverage of the incident.
Key details
- Yglesias recounts preparing personally — locating a household emergency box and an elastomeric respirator he first saw recommended by a biosecurity expert on a podcast in October — and notes that the respirator company recently went out of business, illustrating market fragility for personal protective gear.
- He argues that while consulted medical professionals are not highly worried and public messaging urges people not to panic, structural trends (population growth, increased connectivity, rising prosperity) and advances in biotechnology are raising both natural and engineered pandemic risks, and he criticizes U.S. government under‑reaction and calls for more aggressive countermeasures.
Brief
Matthew Yglesias uses a May 11, 2026 piece to treat the recent hantavirus episodes aboard the cruise ship MV Hondius — reported as five cases and three deaths — as a reminder of broader pandemic preparedness failures. He describes personal steps (locating an emergency box and a specific elastomeric respirator first recommended on a podcast in October) and notes the vendor has since gone out of business, signaling weak consumer supply chains for PPE. Although clinicians he consulted are “not particularly worried,” Yglesias worries about reflexive reassurances and systemic complacency. He emphasizes that demographic and connectivity trends increase the frequency of novel outbreaks, and that biotechnology advances raise engineered‑pathogen risk, concluding that the American government has under‑reacted and that aggressive preemptive measures are warranted.
By Matthew Yglesias