PGT-P seems like generally a good idea if you are doing IVF anyway, but IVF itself introduces significant risks to your baby. Would love to see a blended analysis that balances benefits from screening with risks from IVF for these elective IVF families.

Anyone pumping these bags should go back and take a remedial heat transfer course from their local Chemical Engineering department.

Someone needs to invent tf-idf for new year’s predictions. You don’t get credit for your correct prediction if tf is too high!

Fisher-Price is dooming the next generation of geometers.

Tinker + Claude Code/Codex are nearing the takeoff point where AI can do (or at least verify) AI research. In 30 minutes I reproduced the archaic bird names experiment. This would have taken weeks 6 months ago.

This Thinky drama is quite ill-timed. I just started using their product yesterday and it's great!

Turns out someone did tell him no.

Qwen audio team is on fire. SOTA on ASR benchmarks + cracked singing! The latter is one of the trickiest problems in speech. If someone isn't racing to wrap their models in an enterprise API and take a slice of 11Labs' $10B+ market cap, capitalism is dead.

The Anthropic marketing team is 🔥 "I'm having trouble connecting with mom" "Head to Golden Encounters, the mature dating site that connects sensitive cubs with warring cougars." I gave up this company for dead last year and now I spend $200/day and they make me lmfao.

My 🌶️ take on vibe coding for PMs PMs shouldn’t waste time landing prod diffs at Meta/Google/your-BigCo-here scale. Do it for the love of the game. (and learning and comms and UXR). ddmckinnon.com/2026/02/11/my-…

What do I get as a Gemini Pro subscriber? Stale models and unusable coding agents? This experience needs to be more differentiated from free to be worth the money. Give me the good stuff. — a guy who subscribes to 4 (maybe soon 3) different AI models

TIL Claude Code has 7 commits to OpenAI's OSS Codex repo.

Nice work on the benchmarks from Google, but I am a bit skeptical after finding Gemini 3 underperformed essentially all peers including top Chinese OSS models for my real-world use-cases. How are people using Gemini in coding agents? Jules and Gemini CLI are sooooo far behind.

lol Gemini 3.1 tells me that there is no way to use the model with a functional coding agent.

I thought Google had an infrastructure advantage? I'm no OpenAI stan, but I don't recall anything like this when Codex-5.3 dropped (or Opus 4.6 or Kimi K.25 or ...).

Updated rankings: Codex ~= Claude Code >> Kimi Code > Gemini 3.1 + Antigravity >>> Gemini 3.0 + Gemini CLI. Gemini 3.0 + CLI was unusable as a coding agent. Google has upped its game but still falls short (on my romantasy slop generator and others) despite a strong scorecard

My 🌶️ take on vibe coding for PMs PMs shouldn’t waste time landing prod diffs at Meta/Google/your-BigCo-here scale * If the feature is actually important, fix the system for prioritization (your real job) rather than circumventing it. * You probably code like a slow E3 but cost

In light of the recent Anthropic distillation complaint, it seems timely to share my experience with Chinese coding plans. Codex > Claude Code >> Kimi >> Gemini > Minimax > GLM * Kimi is the best in both quality and harness polish. * The main value seems to be splitting API