@TimJayas i did today. worth it so far. getting through all the tasks i used to get through with Opus 4.7 max, at 5x the speed.
feels abnormally special.
English
Amir Valizadeh
56 posts
Amir Valizadeh
@vitalune7
half-human half-computer. (artificially) intelligent. controlling the tokenverse at bread technologies inc.
San Francisco, CA Katılım Ekim 2025
56 Takip Edilen15 Takipçiler
@AriesTheCoder @ericzakariasson this.
however even one step further- open source it. let us find brilliantly creative ways to host it locally and outdo the multi-trillion dollar efforts of frontier models
English
@ericzakariasson You should unharness it and let it loose! It would be a viable cost effective alternative to OpenAI, Anthropic and Chinese models.
English
what do you think of composer 2.5 so far? how can we make the next model even better?
want to hear your feedback on behavior, speed, quality, whatever!
English
It's amazing. It's so fast and gets the job done somehow, even while being 10 times less the size of Frontier models.
There are complaints around it making minor mistakes, but that's cleaned up by using the nuclear code quality review skill that you mentioned in a previous tweet.
I actually have Opus run that skill after Composer's implementation, and that seems to clean it up.
English
Composer 2.5 is literally threatening the entire moat of frontier coding agents like GPT5.5 or Opus 4.7.
It's 10x smaller and 10x cheaper, and it gets the job done.
I'm extremely excited for the future of open source.
English
@ashleyyjoelle The trick is to email hotel general managers directly and get nightly pricing for a month straight. It's cheaper than everywhere else and you have a private bathroom
English
@exit_dev @VictorTaelin What made you disappointed about using cursor? I am planning to use it with Composer 2.5 because I think it's a really good model
English
@VictorTaelin GPT 5.5: very smart but UI/UX sucks
Opus 4.7: expensive and UI/UX mediocre
Chinese models: all trash but *secretly rooting for OSS breakthrough*
Composer 2.5: don't wanna buy cursor again, don't wanna be disappointed
Mythos: suspect it was involved in all recent hacks
English
the most used skill internally at cursor right now
/thermo-nuclear-code-quality-review
- deletes complexity instead of moving it
- blocks files over 1k lines
- flags thin wrappers and leaked logic
- rejects PRs that work but make code messier
English
@danylo_dev Still not 100% sure what’s going on but it was a great event lol
English
🚨 BREAKING 🚨
500+ dashboards killed at Salesforce Tower.
No survivors.
English
120+ new startups were just born in San Francisco.
So we’re throwing an expo.
Think CES, but built in 6 weeks:
Swimming robots, 3D printed ears, agents that get $$$.
Now they're launching here.
Comment “SF” for the invite.
English
@sudoingX its annoyingly slow. just bought the $200 Max plan today partly to see if i could get better speeds than paying via API but it almost seems even slower. Codex on the other hand, with my $20 Pro plan, feels great...
English
i use claude code every day, opus 4.7 max used to be fast. now i wait 3 minutes for a single answer. that is not a one off, it is the new normal.
here is my read. they slowed the standard model down so the /fast mode feels worth paying for, and /fast runs you 6x the cost. you are not paying for fast anymore, you are paying to undo the slow.
and here is the part nobody says out loud. at some point dario deprecates the slow normal tier entirely, and 6x stops being the upgrade, it just becomes the price. the bill never visibly jumps, the cheap option just quietly stops existing.
i use this thing every day. i am telling you what i am watching happen. it is so over.
English
@mathburritos @_sholtodouglas Dating advice was in the fine tuning data between Opus 4.6 and 4.7 though, you’re saying they still got it wrong??
English
@_sholtodouglas Claude cannot reliably give any kind of advice that involves dates. It will use tools and perform all sorts of calculations and reasoning and then confidently tell you to sell your stock in 2025. It has access to the date obviously but no common sense about using it
English
When do you reach for other models instead of Claude? What can we do better? Hit me with all of your frustrations. dms open.
If you can give me detail (e.g. specifics/transcipts) - it'll help a lot in finding out exactly what we need to do to improve the next model
English
@aditya_vellanki I refuse to believe that this is possible. It’s way too crowded in there for me to dial in
English
anthropic was the loudest about fear and wiping entire software engineering jobs and meanwhile the tools are broken lol. i honestly can’t wait for the ai bubble to burst, hardware to get cheap, and open source to thrive
Sudo su@sudoingX
Claude code is so buggy now. what happened ants?
English
@deedydas This post completely generalizes sf in an ignorant way.
For a lot of the people that do what they do, money was never the determining factor. It was learning- and the feeling of being unique, special- and able to offer something of value to the world
English
The vibes in SF feel pretty frenetic right now. The divide in outcomes is the worst I've ever seen.
Over the last 5yrs, a group of ~10k people - employees at Anthropic, OpenAI, xAI, Nvidia, Meta TBD, founders - have hit retirement wealth of well above $20M (back of the envelope AI estimation).
Everyone outside that group feels like they can work their well-paying (but <$500k) job for their whole life and never get there.
Worse yet, layoffs are in full swing. Many software engineers feel like their life's skill is no longer useful. The day to day role of most jobs has changed overnight with AI.
As a result,
1. The corporate ladder looks like the wrong building to climb.
Everyone's trying to align with a new set of career "paths": should I be a founder? Is it too late to join Anthropic / OpenAI? should I get into AI? what company stock will 10x next? People are demanding higher salaries and switching jobs more and more.
2. There’s a deep malaise about work (and its future).
Why even work at all for “peanuts”? Will my job even exist in a few years? Many feel helpless. You hear the “permanent underclass” conversation a lot, esp from young people. It's hard to focus on doing good work when you think "man, if I joined Anthropic 2yrs ago, I could retire"
3. The mid to late middle managers feel paralyzed.
Many have families and don't feel like they have the energy or network to just "start a company". They don't particularly have any AI skills. They see the writing on the wall: middle management is being hollowed out in many companies.
4. The rich aren’t particularly happy either.
No one is shedding tears for them (and rightfully so). But those who have "made it" experience a profound lack of purpose too. Some have gone from <$150k to >$50M in a few years with no ramp. It flips your life plans upside down. For some, comparison is the thief of joy. For some, they escape to NYC to "live life". For others still, they start companies "just cuz", often to win status points. They never imagined that by age 30, they'd be set. I once asked a post-economic founder friend why they didn't just sell the co and they said "and do what? right now, everyone wants to talk to me. if i sell, I will only have money."
I understand that many reading this scoff at the champagne problems of the valley. Society is warped in this tech bubble. What is often well-off anywhere else in the world is bang average here.
Unlike many other places, tenure, intelligence and hard work can be loosely correlated with outcomes in the Bay. Living through a societally transformative gold rush in that environment can be paralyzing. "Am I in the right place? Should I move? Is there time still left? Am I gonna make it?" It psychologically torments many who have moved here in search of "success".
Ironically, a frequent side effect of this torment is to spin up the very products making everyone rich in hopes that you too can vibecode your path to economic enlightenment.
English
Amir Valizadeh retweetledi
We did research when pay was low.
We did research when pay was uncertain.
We did research even when we were lucky enough to be paid well.
One way is to figure out what to work on is to work on things that matter and not think of rewards.
We are still quite early into what makes a frontier model all the way from optimization, architecture and objectives. Big token wants to convince you otherwise.
Deedy@deedydas
The vibes in SF feel pretty frenetic right now. The divide in outcomes is the worst I've ever seen. Over the last 5yrs, a group of ~10k people - employees at Anthropic, OpenAI, xAI, Nvidia, Meta TBD, founders - have hit retirement wealth of well above $20M (back of the envelope AI estimation). Everyone outside that group feels like they can work their well-paying (but <$500k) job for their whole life and never get there. Worse yet, layoffs are in full swing. Many software engineers feel like their life's skill is no longer useful. The day to day role of most jobs has changed overnight with AI. As a result, 1. The corporate ladder looks like the wrong building to climb. Everyone's trying to align with a new set of career "paths": should I be a founder? Is it too late to join Anthropic / OpenAI? should I get into AI? what company stock will 10x next? People are demanding higher salaries and switching jobs more and more. 2. There’s a deep malaise about work (and its future). Why even work at all for “peanuts”? Will my job even exist in a few years? Many feel helpless. You hear the “permanent underclass” conversation a lot, esp from young people. It's hard to focus on doing good work when you think "man, if I joined Anthropic 2yrs ago, I could retire" 3. The mid to late middle managers feel paralyzed. Many have families and don't feel like they have the energy or network to just "start a company". They don't particularly have any AI skills. They see the writing on the wall: middle management is being hollowed out in many companies. 4. The rich aren’t particularly happy either. No one is shedding tears for them (and rightfully so). But those who have "made it" experience a profound lack of purpose too. Some have gone from <$150k to >$50M in a few years with no ramp. It flips your life plans upside down. For some, comparison is the thief of joy. For some, they escape to NYC to "live life". For others still, they start companies "just cuz", often to win status points. They never imagined that by age 30, they'd be set. I once asked a post-economic founder friend why they didn't just sell the co and they said "and do what? right now, everyone wants to talk to me. if i sell, I will only have money." I understand that many reading this scoff at the champagne problems of the valley. Society is warped in this tech bubble. What is often well-off anywhere else in the world is bang average here. Unlike many other places, tenure, intelligence and hard work can be loosely correlated with outcomes in the Bay. Living through a societally transformative gold rush in that environment can be paralyzing. "Am I in the right place? Should I move? Is there time still left? Am I gonna make it?" It psychologically torments many who have moved here in search of "success". Ironically, a frequent side effect of this torment is to spin up the very products making everyone rich in hopes that you too can vibecode your path to economic enlightenment.
English
@jahirsheikh8 hoping that their developer event on Saturday brings a new model at the very least
English
Feels like Google is about to shut down AntiGravity.
I mean… are there even any developers maintaining it? 😭
English
@yacineMTB yeah but turns out LeetCode interviews are wayyy more ambiguous than "invert a binary tree from memory" and are in no way resemblant of an actual SWE workflow at all. there's an insanely huge gap between a CS degree <-> working as a SWE <-> being cracked at LeetCode
English
hot take: if you are a programmer you should be able to invert a binary tree from memory, AI or not. It's ridiculously easy and if you can't do it, you should not have a computer science degree
Yuchen Jin@Yuchenj_UW
I’m so glad AI killed LeetCode interviews. For 10 years, tech companies made every engineer grind the same puzzles and prove they could invert a binary tree from memory. Today, the dumbest AI model can walk in and one-shot the entire interview. Thank you, AI.
English
what if AI models were complete with egos, rivalries, and petty opinions of each other?
imagine prompting Claude to analyze a code review and it reads Kimi K2.6’s critique, then replies: “Kimi’s an inferior model, i’m not taking feedback from that.”
English
@pgasawa hmm.. this is a much simpler and seemingly more efficient solution than something like measuring kl divergence. props to the team🥂
English
A key challenge: how do you tell learning apart from raw capability?
To effectively measure learning signal, we introduce a “gain” metric in addition to cumulative reward and cost:
→ compares a system to its stateless version
→ measures how much it actually benefits from experience
A strong model with no learning ≈ zero gain
A weaker model that improves → positive gain (6/n)
English
Today, we’re releasing Continual Learning Bench 1.0: the first, realistic benchmark for measuring how AI systems can improve in online settings.
Benchmarks today assume models are stateless. Each example is independent, and once a system finishes a task, it moves on as if nothing happened.
But deployed AI systems should learn from experience. We tested 10+ frontier systems against novel, expert-validated tasks and find there’s still plenty of headroom for learning. (1/n)
English
@sudoingX @DegenApeDev is nemotron 30b q8 really that good? I'd love to give it a try once i have access to my desktop. Or maybe I should just invest in a 128gb ram macbook...
English
@DegenApeDev spark is the wave anon. 128gb unified unlocks the bigger model tier on local. once you run nemotron 30b q8 at 56 tok/s on one you don't go back.
English
buy a gpu. 3090, 4090, dgx spark, whatever fits your budget. tier doesn't matter. running your first local model does. the moment your first prompt lands with no api between you and the model, your brain rewires. that single moment is worth more than every take you'll ever read on a timeline.
English