TDM (e/λ) (L8 vibe coder 💫)
@cto_junior
Followers
13K
Following
82K
Media
4K
Statuses
26K
L-12 in ZIRP | Larry Ellison's bloodboy | One-man soonicorn | Softbank & FIITJEEs love child Google doc for my AGI notes ⬇️
Joined April 2022
Enjoy the origin story of Wizard-Vicuna-13B-Uncensored-SuperHOT-8k-GGML models. After this you can also pretend like me to be an AI enthusiast on /lmg
45
200
1K
GPT-5.2 feels sluggish in codex compared to 5.1 either the model got bigger or the deployment is not optimized yet
2
0
10
Is GPT-5.2 the first public model pretrained on Blackwells?
Gavin Baker, Managing Director of Atreides Management, explains how reasoning saved AI progress on @InvestLikeBest. We have yet to see a model trained on Nvidia Blackwell. The massive gains of the past 18 months are due to TTC. There is no wall.
1
2
39
impressive long context perf by gpt 5.2 also touched 80% on swe-bench at par with opus 4.5 at tau-bench-telecom, slightly less in retail one
9
2
24
OpenAI won
1
0
12
Imagine if everything in life was as good as gold. It’s easy to invest in, could help grow your wealth, and protect it too. A timeless asset that stands strong when others may not.
130
243
4K
GPT-5.2 is here and OpenAI has finally lifted the knowledge cutoff curse
1
1
36
Broad world knowledge! @xeophon_ we might finally get past the openai cutoff curse
New Codex PR seems to indicate GPT-5.2 is dropping very soon 👀 "robin" strings have been updated to "gpt-5.2" https://t.co/GuOmQb7IIp
1
0
2
Tuned the workflow toward actual data distributions instead of vague shape metaphors Way cleaner, Way more intuitive
Pipeline zero shotted this voxel simul for avataRL repo by @tokenbender I simply fired the agent with prompt `explain all the steps the training algorithm` then stepped out with my wife for the newest fancy smash burger shop in town came back after 3 hours to find this lovely
1
0
10
System.out.println as a service
Cursor can now fix your hardest bugs. Debug Mode instruments your code, spins up a server to capture logs, and streams runtime data to the agent. Also in 2.2: Plan Mode improvements, multi-agent judging, and more.
1
0
12
Opus 4.5
0
0
3
Quickly and easily evaluate borrowers' income with our award-winning Income Calculator. Let our technology work harder for you, so you can do great work for your borrowers. Learn more.
2
11
115
They're right The threejs sim agent I'm building is basically Codex + detailed guidelines + a starter kit + a tiny MCP that lets you play browser games. It just works. No need to hand roll loops anymore
i was skeptical when @simonw said that "Claude Skills are awesome, maybe a bigger deal than MCP" buuut early indications are this is correct. this is the fastest talk ever to pass 100k views here at AIE. its like those 0 - 100m ARR charts but for attention. @MaheshMurag and
0
0
2
@cto_junior feels like the constraint isn't the language, it's the fact that most 40-year-olds aren't actually trying anymore. kids have energy. that's it. the language they use is mostly irrelevant to whether they'll build anything meaningful
0
1
2
The next decade belongs to these kids so build AI products that speak their language instead of ours
Watching my 9-year old code on top of Claude makes it clear that he is using a new programming language. It has a weird input structure. He draws diagrams (or app states / flows) and then reasons about it. He then iterates on it before making new diagrams. It’s a multi-modal
1
0
6
My moat? Always reading the second tweet in the thread that contains the link to the alpha
2
1
20
One dumb thing Codex does is obsess over backward compatibility even when you ask it to tweak a feature it wrote yesterday
2
0
11
Pipeline zero shotted this voxel simul for avataRL repo by @tokenbender I simply fired the agent with prompt `explain all the steps the training algorithm` then stepped out with my wife for the newest fancy smash burger shop in town came back after 3 hours to find this lovely
Workflow looking more robust now It’s now able to run for more than 2-3 hours by itself carefully crafting each detail in animations only remaining hiccup right now is grounding shapes of voxels with some real world object
3
6
35
it’ll never stop being funny to me that statisticians got mogged by some CS people who were like “what if we just split everything like a tree and did that over and over to predict anything?”
35
50
3K
Train smarter, store freely - no egress fees on Cloudflare's R2 object storage.
5
14
171
releasing v0.1 of Papercode today! - a fun platform for getting back to implementing papers from scratch - you can implement papers from scratch using leetcode style interfaces :D - auth, good storage coming soon - this is a barebones structure at this point, so please report
38
44
621
Presenting gpt-5.1-codex-max
GPT-5.1 codes like a paranoid L3, Opus-4.5 like a git commit maxxing L4, and Gemini-3.0 cooking like an L5 artisan
0
0
16