Articles by gmays
1

How the World Became a Casino (podcasts.apple.com)

1

Agentic Search Models (softwaredoug.com)

2

Supply Chain Inheritance (citriniresearch.com)

5

Build iterative repair loops with Codex (developers.openai.com)

3

Scientists ID 'corkscrew killer' behind gruesome seal deaths (science.org)

2

The Interstitium, the Human Body's Hidden Pathways (nytimes.com)

6

Elon Musk Announces xAI Will Become SpaceXAI Division (notateslaapp.com)

8

Localmaxxing (tomtunguz.com)

5

A²RD: Agentic Autoregressive Diffusion for Long Video Consistency (dxlong2000.github.io)

2

The Inference Shift (stratechery.com)

2

The Main Path to Creative AI (danielmiessler.com)

2

Normalizing Trajectory Models (arxiv.org)

1

A Fish That Hitches Rides Where the Sun Doesn't Shine (nytimes.com)

10

Texas county pauses data center construction in rural areas (texastribune.org)

1

A Single Infusion Could Suppress HIV for Years (nytimes.com)

2

Google says it disrupted an AI-driven effort to exploit a software bug (apnews.com)

1

Running Codex Safely at OpenAI (openai.com)

1

Improving token efficiency in GitHub Agentic Workflows (github.blog)

3

How the Cheesecake Factory Runs One of America's Biggest Menus [video] (youtube.com)

1

Harvey's Legal Agent Benchmark (harvey.ai)

2

Anthropic tests Jupiter-v1-p ahead of its developer conference (testingcatalog.com)

2

OpenAI Flips the Script (every.to/context-window)

3

China to Invest in DeepSeek at $50B Valuation (wsj.com)

1

CyberSecQwen-4B: Why Defensive Cyber Needs Small, Specialized, Local Models (huggingface.co)

1

Dinosaur dental fossils reveal bird-like parental care bonds (osu.edu)

1

Reliance on Middle East Oil and Gas Supplies by Country (iea.org)

1

SFT, RL, and On-Policy Distillation Through a Distributional Lens (nrehiew.github.io)

1

All the demons hiding in your AIs (drtompollak.substack.com)

1

Long AI Short AGI (1984.substack.com)

2

The Cost of Overfitting the Harness (dbreunig.com)

1

World Models Can Change Everything (weightythoughts.com)

1

In-Kernel Broadcast Optimization: Co-Designing Kernels for RecSys Inference (pytorch.org)

3

The Problem with "Mathematically Proven" Claims About LLMs (webdirections.org)

1

How AI Agent Memory Works (cobanov.dev)

1

In a quest to becoming AI-independent (adlrocha.substack.com)

1

Good QC for RL Data (seancai.com)

2

/Goal: The Six-Hour Codex Run That Survived a Five-Hour Pause (tectontide.com)

6

Supercomputer networking to accelerate large scale AI training (openai.com)

1

Tuna-2: Pixel Embeddings Beat Vision Encoders for Understanding, Generation (github.com/facebookresearch)

3

Notes from inside China's AI labs (interconnects.ai)

1

Scientists Gave 'Aggressive' Fish Psychedelic Drugs. A Breakthrough Came Next (404media.co)

1

Motion Picture Editor's Guild Stress Survival Kit (editorsguild.com)

3

The April every AI plan broke (thefinancialengineer.substack.com)

1

When My Father's Canary Flew Away (nytimes.com)

2

Meta-Meta-Prompting: The Secret to Making AI Agents Work (twitter.com/garrytan)

1

AI inference just plays by different rules (theregister.com)

9

Fossils show millipede and centipede ancestors evolved legs underwater (phys.org)

3

Google updates AI search to include quotes from Reddit and other sources (techcrunch.com)

5

Nvidia embraces AI investor, topping $40B in equity bets 2026 (cnbc.com)

2

EMO: Pretraining mixture of experts for emergent modularity (allenai.org)

2

The Anti-Singularity (lesswrong.com)

1

Why MistralAI Grows Faster Than OpenAI/Anthropic (productify.substack.com)

3

SkillOS: Learning Skill Curation for Self-Evolving Agents (arxiv.org)

2

Useful Memories Become Faulty When Continuously Updated by LLMs (dylanzsz.github.io)

1

Oil in Everyday Things (reuters.com)

2

Metformin's real power may be in the gut (northwestern.edu)

2

Unconscious Brain May Process Sound, Learn, and Predict Words Under Anesthesia (discovermagazine.com)

3

New CRISPR Could Treat Viral Infection and Cancer by Shredding Sick Cells' DNA (utah.edu)

2

What can singing mice say about human speech? (phys.org)

3

A man who blew up a nuclear power station and disappeared (theguardian.com)

2

Designing, Refining, and Maintaining Agent Skills at Perplexity (perplexity.ai)

1

Anthropic and OpenAI are launching joint ventures for enterprise AI services (techcrunch.com)

1

In search of wasted bits: how much information do LLM weights carry? (fergusfinn.com)

47

Gemini API File Search is now multimodal (blog.google)

2

Powering the Inference Era: Inside the DigitalOcean AI-Native Cloud (digitalocean.com)

3

Maryland Power Bills: $1.6B Data Center Subsidy (unusualwhales.com)

1

You Are Not Immune to Mode Collapse (lesswrong.com)

1

Hugging Face's Clem Delangue: Stop Comparing Engines to Cars (turingpost.com)

1

Consumer AI's ARPU Problem (twitter.com/sashakaletsky)

21

The context window has been shattered: Subquadratic debuts a 12M token window (thenewstack.io)

1

Deepsec: The security harness for finding vulnerabilities (vercel.com)

2

vLLM Routing and KV (avkcode.github.io)

2

Reduce friction and latency for long-running jobs with Webhooks in Gemini API (twitter.com/googleaistudio)

2

How to Scale Your Model (jax-ml.github.io)

2

Meta plans advanced 'agentic' AI assistant for users (reuters.com)

1

Model-Harness-Fit (twitter.com/nicbstme)

6

Hallucinations Undermine Trust; Metacognition Is a Way Forward (arxiv.org)

1

Apple plans to make iOS 27 a Choose Your Own Adventure of AI models (techcrunch.com)

2

Meta will use AI to analyze height and bone structure to identify underage users (techcrunch.com)

1

End-to-End Autoregressive Image Generation with 1D Semantic Tokenizer (arxiv.org)

1

How LLM Inference Works (twitter.com/akshay_pachaar)

1

Replit's Amjad Masad on Cursor deal, Apple fight, and why he'd rather not sell (techcrunch.com)

1

Coding plan comparisons based on actual usage (sites.diy)

55

GPT-5.5 Price Increase: What It Costs (openrouter.ai)

4

Automating AI Research (jack-clark.net)

5

LLMs Distort Our Written Language (sites.google.com)

2

Anthropic working on Orbit, its upcoming proactive assistant (testingcatalog.com)

1

Synthetic Computers at Scale for Long-Horizon Productivity Simulation (arxiv.org)

2

Researchers pioneer method to rapidly design proteins (stanford.edu)

3

A small object past Pluto may have a thin atmosphere (sciencenews.org)

2

How did 'large' language models get that way? (greaterwrong.com)

1

Yellowstone's volcano may be fueled differently than previously thought (livescience.com)

9

Google to sell TPU chips to select customers (yahoo.com)

1

Batch API is terrible for one agent. It might be great for a fleet (sandler.co.il)

1

Climate change impacts are making winters shorter (usatoday.com)

3

Recursive Language Models, clearly explained (twitter.com/akshay_pachaar)

1

What you're writing when you write a SKILL.md (laxmena.com)

1

KV Cache Locality: The Hidden Variable in Your LLM Serving Cost (ranvier.systems)

1

Granite 4.1 LLMs: How They're Built (huggingface.co)

1

Ford factory workers get 40-hour week (history.com)