graadient

･ *ﾟ･ descent 　 ･ ﾟ* ･｡ *･｡ *.｡ ｡･ °*. ﾟ ﾟ*. ｡｡ ･ ｡ ･ﾟ ｡°*. .｡*･｡

@azaztrader01 to boldly grow where no man has grown before

@graadient I see you're using advanced math to reach more users than exist. Bold strategy.

Little known growth-hack. When publishing your app to the AppStore/Playstore. Just set the roll-out percentage to a number higher than 100%

@dwr One key consideration some builders mix up is to not consider a passkey as the canonical/stable identity but rather as a substitute for an ephemeral password. We did that mistake with Zeal initially

Passkeys are good! Spent the last few years using them in production. There are definitely footguns -- and they are not a panacea -- but if properly scoped and you understand the trade-offs, they can make for great UX for most people.

@levelsio *bubble up some internal info to the menu

@levelsio This and a tmux session menu when logging into shell so I can browse and select between active sessions. Tmux can bubble up some intervals to the key so you get a nice overview

Are you guys aware I am coding mostly on my phone now all day via Termius to Claude Code on my server while I go with gf to the dentist, clothing store, cafe, etc. 😛✌️

so close now...

it's not called a Millennium problem for nothing

Please solve Navier-Stokes. Make no mistakes. 1470m 16s and still grinding

@_lewtun This is me doing math

We trained a tiny 4B model to reason for millions of tokens through IMO-level problems. Heaps excited to share our new blog post covering the full pipeline, from distilling the 🐳 to augmenting RL with a reasoning cache that unlocks extreme inference-time scaling for theorem proving. huggingface.co/spaces/lm-prov…

@nikitabier People may think this is a joke

My agent looked up every Amazon product I've bought in the last 10 years, called each manufacturer, said it broke and demanded a replacement. I now have 6 TVs, 12 printers, 2 microwaves, and 800 tubes of tooth paste.

Benchmarking duration of agentic workstreams is a bit like trying to maximise time to get to destination

Trick for better Codex implementation plans; tell your planner to write the plan for an autistic engineer

@banteg @EthanLipnik wow... cutover really does do wonders. Codex is bringing out the scythe on the codebase

@EthanLipnik the 5.3 word for it is cutover

This is the most annoying part of codex. Every prompt I have to say “do not worry about backwards compatibility”

Still going strong

@OVGNFT Sorry, I will stop apologizing

@graadient Stop apologizing and just be right, Claude😭

Anthropic Opus; strong beliefs loosely held. - X is absolutely right! Are you sure? - Actually, y is correct! Are you sure? - No, sorry x is actually correct!

@LyalinDotCom @Angaisb_ Not seeing great performance on hard math from new DeepThink. Advice on best practices working with it and lean?

@Angaisb_ It’s really good

I can't wait for Gemini 3 Deep Think in the API I need to try itttt

@aakashgupta Most insightful comment on the psychology of ai-assisted coding I’ve read

This is a dopamine loop, and it’s one of the most powerful ones humans have ever encountered. Every time you prompt an AI and get a useful result back in seconds, your brain gets a hit. Variable-ratio reinforcement, same mechanism as slot machines, except the reward is real: actual output, actual progress, actual leverage on your ideas. Traditional work follows a delayed-reward structure. You write code for 6 hours, maybe it compiles, maybe you get feedback in a week. The gap between effort and reward is wide enough that motivation decays constantly. AI compresses that loop to seconds. Effort → reward → effort → reward. Your prefrontal cortex stays engaged because the next payoff is always one prompt away. This is why people describe it as “fun” when they’re actually working 14-hour days. The subjective experience of effort disappears when reward frequency is high enough. The “harder than ever” part is real too. When your bottleneck shifts from execution to imagination, you run out of excuses to stop. There’s no “waiting on the build” or “blocked by review.” Every idea you have can be tested immediately, which means your brain never gets a natural stopping point. People who thrive on this are selecting for a specific neurotype: high novelty-seeking, high conscientiousness, tolerance for rapid context-switching. That’s maybe 10-15% of the population. The other 85% will experience the same tools as overwhelming, not energizing. And that split is going to define the next decade of who captures value from AI and who gets displaced by it.

Codex 5.3 is surprisingly good at doing Lean4 proofs when forced, although it loves to scaffold around problems instead

@Dominanski @YinYangTurbo @woke8yearold Feels like we’re close to the point where you can pick your favourite sci-fi novel and “just do it”

@YinYangTurbo @woke8yearold I think I picked it up from science fiction and computer games, then reading kurzweil and others served as confirmation. Have a childhood friend who got it hey @graadient but almost everyone else was dismissive...

It’s super weird that I always thought the singularity would happen but now it actually is Knowing something in the mind is different from knowing it in the gut