AI tinkerer | Building AI agents for finance and personal productivity | Prev: built voice/video calling @ WhatsApp
Joined March 2008
- Tweets2K
- Following511
- Followers244
- Likes1.7K
Manpreet Singh retweeted
Cavemen Issues Dire Warning That The Wheel Is Going To Eliminate Countless Jobs
Astra is a really good model. But I don't know how @OpenAI justifies the $500 price tag given the $200 Claude Code subscription.
The GPT-6 sol and luna models from @openai are remarkable in the cost and performance combination they offer. They have been able to stay ahead of all open source models when you compare on a cost/performance efficient frontier. @GoogleDeepMind previously did a great job at this.
Manpreet Singh retweeted
“Don’t bend; don’t water it down; don’t try to make it logical; don’t edit your own soul according to the fashion. Rather, follow your most intense obsessions mercilessly.” — Franz Kafka
It seems people calling out chinese models like kimi k3 as built with distillation is just cope.
The sooner the US model labs realize that open source poses competition and plan accordingly, the better.
FWIW, I think gpt 5.6 models have great price/performance even when to bleeding edge open source. @AnthropicAI needs to get more competitive.
Both US labs seem safe on the coding side. Only the very best coding models will have max adoption not tier 2.
If you look at the coding use cases, the new kimi k3 model is second only to gpt 5.6. According to livebench at least. Curious to see on DeepSWE but they haven’t updated yet.
Manpreet Singh retweeted
you think you need to “push yourself.” how unwise you are. have you ever tried being pulled? do you know the definition of wisdom? wisdom is sending all your blood to your brain all the time to keep alive an image of you so glorious that not moving day and night toward it is instant suicide. get dragged.
Missed the last 10mins of the England vs Argentina game, for a meeting, thinking "it's done" and was surprised as hell when I came back and saw the score at 2-1. And Argentina had won! Amazing game!
Looking forward to Argentina vs Spain on Sunday.
Would be amazing to see if @thinkymachines catches up to current SOTA open source LLMs and we get home grown abundant and cheap AI. Today's release was a nice leap.
codex is seeing rapid increase in usage and getting somewhat slower like how Claude Code started to become when it shot up in popularity past Opus 4.5.
@OpenAI keep up the speed of gpt 5.6 and keep it usable. It's definitely getting slower by the day.
GPT 5.6 sol clearly improves over my previous coding workhorse GPT 5.5. It find some obscure bugs in my codebase when I switched to it, something GPT 5.5 hadn't found. Which is saying something...
Codex removed the 5-hour limit. Amazing! Really hope they make this permanent. One weekly limit FTW!
Bold moves by @Google launching a new AI-powered device: googlebook.google/
@Apple this was supposed to be your move.
Manpreet Singh retweeted
It’s criminal how cheap and how good Gemini Flash is.. that too with 1M context windows and structured outputs.
Probably, my most used model in production workloads.
Separately their new live voice model is mindblowingly good if you haven’t tried it yet!
Manpreet Singh retweeted
Literally just having a delusional golden retriever mindset measurably changes outcomes and physiology.
Sleep badly? Convince yourself you're well rested.
Stressful day? Convince yourself it's fuel.
Failed? Convince yourself it's useful data.
Manpreet Singh retweeted
go to bed
right now
i know the build is almost finished
the eval can wait til morning
the agent will still be failing tomorrow
you won't figure out why it's hallucinating
yes your coworker ships on 4 hrs of sleep
they also hallucinate a lot
off you go
Manpreet Singh retweeted
some random ai thoughts:
- for code, i went from 80/20 claude/gpt to 80/20 gpt/claude in <3 months. surprised by this tbh, and interested to see where the split is at in another 3mo.
- claude still mogs gpt for non-coding agent stuff. codex feels like an engineer (which is great for coding!), whereas claude still feels like a general purpose coworker. gpt still lacks that coworker magic
- i’m pretty meh on opus 4.7. my experience hasn’t been *bad*, but it certainly hasn’t been good. sideways if anything.
- anthropic has got to figure out the compute thing. you can feel it as a user. vibes are all out of whack bc of it. my opinions above are all likely downstream of this. it’s an issue.
- anthropic labs continues to be the goat of ai product. claude design is another hit. it’s fantastic. idk why it’s not talked about more? a+
- updated claude code app is great. i finally switched out of the terminal for it. very well done.
- how are people STILL sleeping on the claude agent sdk? i feel like i’m going insane.
- gpt 5.5 is incredible. the level to which i trust it for engineering is amazing. if i could only have one model rn, it would be this one just bc of strong need for the coding use case.
- codex team is killing it. app has been the gold standard since 5.3 release (buuut i credit conductor team for the ui innovation that everyone is using now). though i could do with a little less passive aggressive shots at ant from the codex team. TARS, dial up class by 30%. it’s a long race guys haha
- i uninstalled cursor this month and am now back to vs code for my ide. composer just can’t hang with claude/gpt, and the product feels a bit all over the place. pretty stoked about the xai thing though, because their team is absolutely stacked and i’m excited to see what they might be able to do with that compute. codex and claude code are t1, cursor is t2. i would love if this deal got xai/cursor to t1 for a real trio there.
- gemini…? seems like this is 2-3 models now where the model seems like a great release and then nobody ever uses it? i’m bullish google/deepmind but weird it hasn’t translated to product use in any form. kinda disappointed still
- no open source models have hit the opus 4.5 level. was hopeful the new deepseek would get there, but nope. good oss agents will have to wait a few more months it would seem…