Tyranny is the deliberate removal of nuance.
Joined April 2008
- Tweets28.1K
- Following431
- Followers2.8K
- Likes6.6K
I’ve conceded to the fact that I’m just carrying a iPad mini around now. Don’t even bother with folding it.
211
We need a new benchmark. I propose the name NeurosurgeonBench.
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math.
It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training.
We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵
Blog: rikiyat.github.io/blog/weigh…
Readers added context they thought people might want to know
LLM was given weights explicitly as text in prompt and used code/forward passes (per authors' logs) to ablate Qwen looping direction, a known thinking-mode issue per model card. Primarily fixed repetition in that mode, not general 'natively' improving weights/math.
rikiyat.github.io/blog/weight-sp…
github.com/RikiyaT/lm_und…
huggingface.co/Qwen/Qwen3.5-0…
huggingface.co/RikiyaT/Qwen3.…
1
229
Claude Code desktop's session list has a triple dot button. Clicking on it reveals a context menu with actions for the thread. Each action has a keyboard shortcut. But they only work if the context menu is showing 🤔
1
176
Code is your literal source of truth. Docs and tests should only exist if they don't repeat information in code.
A prompt for native app builders.
Go through all my unit tests and identify tests or assertions matching the following:
* realitically relies on the UI frameworks, and the assertion would break if the framework behavior changes, and there'd be no actionable step to fix it other than updating the test again?
* renders the UI but in reality it's asserting on results of a value being rendered, therefore should be replaced by headless tests on the alue being correctly delivered?
* restates constants, prompt strings, table rows in the app logic?
* repeats logic from app's implementation, then assert on the output of said logic with no calls to the actual app code?
Models get better at hacking.
More private data leaks.
New models train on private data.
Models get better at hacking.
RSI.
Bed time reading
The Rust style guide used by Google has been open sourced!
google.github.io/styleguide/…
4
2,839
Was pretty bad. Reminds me of trying to get out of EDC
San Francisco drivers who suffer lingering trauma from last month’s epic traffic jam in Golden Gate Park might consider lying low this weekend. sfchronicle.com/sf/article/h…
1
1
390
Daniel Duan retweeted
Hi future coworker 👋 We're hiring a UI Frameworks Evangelist to join the team. SwiftUI people this is your timeeeee.
jobs.apple.com/en-us/details…
