RESEARCH

–

16 April 2026 19h55

MRCR is a bad eval, we use GraphWalks now. More info here:

→ View original post on X — @simonw

Qwen 3.6-35B-A3B vs Claude Opus 4.7 SVG Generation Comparison

By

@simonw

–

16 April 2026 19h42

Here's Qwen 3.6-35B-A3B v.s. Claude Opus 4.7 for "Generate an SVG of a flamingo riding a unicycle", in case you thought Qwen might be cheating at the pelican benchmark

→ View original post on X — @_jasonwei

AI Transforming Healthcare Through Comprehensive Patient Data

By

@_jasonwei

–

16 April 2026 19h37

Beautifully written piece by @FAbnousi about how AI for health might look like in the future
The current data in health is limited because it only captures episodic clinical snapshots of what happens to our bodies
The revelation is that there is so much latent knowledge in

Letting Models Think: From Manual Temperature Control to Automatic Decision-Making

By

–

16 April 2026 19h34

Better to let the model think. It's sort of like how we used to manually set "temperature" two years ago — nowadays it's better to let the model decide.

→ View original post on X — @swyx

Mid and Post Training Compute Now Rivals Pretraining Cost

By

@swyx

–

16 April 2026 19h33

i think in a world where mid and post training take equivalent or more compute (!!) than pretraining, this is less big a deal than what it used to mean in 2023

→ View original post on X — @simonw

Local Qwen 35B Outperforms Claude Opus 4.7 on Benchmark

By

@simonw

–

16 April 2026 19h27

Shocking result on my pelican benchmark this morning, I got a better pelican from a 21GB local Qwen3.6-35B-A3B running on my laptop than I did from the new Opus 4.7! Qwen on the left, Opus on the right

Phasing Out Flawed Evaluation with System Card Caveat

By

–

16 April 2026 19h26

This is a bad eval that we've been phasing out. Going to add a caveat to the system card to make it clear. More here:

→ View original post on X — @mit_csail

MIT CSAIL Harvard AI Accelerates Electron Microscope Imaging Tenfold

By

@mit_csail

–

16 April 2026 19h22

A decade of imaging.

Compressed into three months.

Here’s how MIT CSAIL & Harvard taught an electron microscope to see like you do: https://t.co/qC3zVck0HF pic.twitter.com/kMTptacRMD
— MIT CSAIL (@MIT_CSAIL) 16 avril 2026

A decade of imaging. Compressed into three months. Here’s how MIT CSAIL & Harvard taught an electron microscope to see like you do: https://
bit.ly/4tfPxBX

MRCR Phase-Out: Shifting from Distractor-Based to Applied Long-Context

By

–

16 April 2026 19h12

We kept MRCR in the system card for scientific honesty, but we've actually been phasing it out slowly. Two reasons: (1) it's built around stacking distractors to trick the model, which isn't how people actually use long context, and (2) we care more about applied long-context

Phasing Out MRCR: Shifting Focus from Distraction Tricks to Applied Long Context

By

–

16 April 2026 19h10

We kept MRCR in the system card for scientific honesty, but we've actually been phasing it out slowly. Two reasons: (1) it's built around stacking distractors to trick the model, which isn't how people actually use long context, and (2) we care more about applied