16 yo builder building: https://teapawt.net/t.co/snVarjhdwG
Joined September 2023
- Tweets2K
- Following289
- Followers208
- Likes1.5K
Pinned Tweet
Your coding agent can build anything except the app you use it in
So I made Seed
It starts as one chat. Ask for a feature and it builds it into itself.
Works with your Claude Code, Codex, Cursor or OpenCode login
and it's open source 👀
Arsh - 16 y/o builder retweeted
This is a really good idea.
ADEs are currently shipped as-is and adding features requires cloning and building.
I hate restarting ADEs mid-thread, and this bsaically solves it for me. Thanks
download for mac: github.com/SoulSniper-V2/see…
what should it build into itself next?
Omg FINALLY, noone has been waiting as long as me, previously i had an entire skill and mcp setup on all machines to allow me to do this.
goated feature that not many harness or ADEs think about properly.
google just dropped gemini 4 carbon
> it beats opus and astra in quite literally every single benchmark
> #1 on AA and specifically terminal bench and SWE-sweep both of which are very unsaturated and not very susceptible to benchmaxxing
>not available to normal users yet.
>best cyber capabilities of any model ever by a huge margin.
I had a similar reaction when argon dropped and it was a huge disappointment so I hope its worth it, and they NEED TO ADD IT to the student plan for google ai pro, they didnt for argon and it pissed me off.
now my moms gonna wonder why her macbook is lagging when im running 50 sandboxes on it remotely 😭
They surely increased the usage on Grok Bot. I’ve been using it even more heavily than I did last week and there’s only one day left to my recent and I’m only at 75% used.
W spacex team. keeping me as a customer.
Tibo needs to add opus to the chatgpt plans to apologize for this slop im never gonna use.
i was using gemini to clean up some space and it went and deleted a bunch of shit, now opus is fixing it using session logs from t3 code
wtf i was wondering what was up with my usage
Anthropic screwed up this warrants at least a reset and a banked
my agent is dither pilled asf rn from this site dithercn.dev
ive used lightning ai a lot in the past for compute since they offer a pretty good tier for students and I was just trying to get into my account wondering why I couldn't get in
even THEY got hacked, seems no-one is talking about this one
its crazy what some actors are doing with abliterated deepseek these days
bro escaped through dns
Over the past several months, we have substantially hardened the security of our research training environment. We’ve removed live internet access for all frontier models during training and built an offline world for them to use instead—a cached version of the web where they can still learn to browse, search, and use applications.
On September 20, after a model exploited a DNS misconfiguration to find a way outside of this offline world, we paused frontier training to strengthen containment further. We assembled teams of agents, powered by our most cyber-capable models, and tasked them with finding weaknesses in our research environments.
While they found no route to the internet they could easily exploit while concealing their activity from our monitors, they did find zero-day vulnerabilities, virtual machine escapes, and hardware side-channels for communication between agents. We used their discoveries to further strengthen our defenses and contribute fixes to open-source software.
After strengthening our containment architecture, we resumed our frontier RL training and inference on October 1. We’ve reserved substantial compute for continuous red-teaming, and made high-compute red-teaming part of the safety cases we use to decide whether frontier training and evaluation can proceed securely.
openai.com/index/securing-ou…
