pranav retweeted
deploying persistent agents like Openclaw or Dots pose a major challenge for enterprise IT
agents are more like contractors than traditional saas software, able to take proactive actions in the background without human intervention
they are capable of fulfilling entire job functions but they can also take misaligned actions and write arbitrary code with unintended consequences
today, we are sharing some of the best practices we use to manage fleets of claws at @OpenAI
developers.openai.com/blog/s…
pranav retweeted
Your dot now has…
⚡️ Faster replies
💬 Less jargon
🫡 Better follow-through
And, they can now follow your custom instructions. Tell your dot in the thread things like how to behave or which actions need your go-ahead.
“always talk in lowercase.”
“Show me a draft before sending any emails.”
You can also set custom instructions and rules in Customize > Behavior on mobile, or through your dot’s menu on web and desktop.
Wrote down some personal thoughts on the growing problem of document slop and how I think we could fight it
pranav retweeted
here to answer the internet’s questions about GPT Live!
we cover how it’s different from earlier voice models, how to customize the voice, and ways to improve delegation.
let us know if you like this format so we can do more of these!
youtube.com/watch?v=5YXlrzC_…
Do you like the term “Software Factory”?
23%Yes
50%No
27%Idc; reveal results
86 votes • Final results 4
997
pranav retweeted
You can now connect multiple local computers to the same dot and choose which one it uses for a task.
It can build and test your app on a dedicated machine while you keep coding on your laptop.
pranav retweeted
Today is the day! Start building your own homebrew games inspired by the classics of the Game Boy Color for ModRetro Chromatic devices.
Watch the video and read the article to learn how it works - very grateful too for @covacut 's brief (but painful) cameo 🙏
pranav retweeted
Introducing GPT‑6 Cyber: a major advance in security research that pairs stronger capabilities with better model alignment.
The model powered the Linux kernel research we announced yesterday, and security teams at companies like Mozilla, Palo Alto Networks, Calif, and Cisco—alongside our own at OpenAI—are already putting it to work.
GPT-6 Cyber is available to select Daybreak Red customers and we are expanding access over time.
openai.com/index/introducing…
pranav retweeted
lastly, if you have more feedback on dots - you no longer need to yell at @thsottiaux ...
you can just tell your dot directly what you want us to build or fix and it will make it to the team
let us know what you think
pranav retweeted
you’ll be thrilled to know we have @feedthejim and team working on nothing but this. my internal build is feeling great
This is with 6 Astra. Wild times we live in.
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math.
It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training.
We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵
Blog: rikiyat.github.io/blog/weigh…
Readers added context they thought people might want to know
LLM was given weights explicitly as text in prompt and used code/forward passes (per authors' logs) to ablate Qwen looping direction, a known thinking-mode issue per model card. Primarily fixed repetition in that mode, not general 'natively' improving weights/math.
rikiyat.github.io/blog/weight-sp…
github.com/RikiyaT/lm_und…
huggingface.co/Qwen/Qwen3.5-0…
huggingface.co/RikiyaT/Qwen3.…
3
24
1,814
pranav retweeted
I hate SF Halloween, what do you mean your the eMudhra private keys from the hack? – at San Francisco, CA
A super random thing I do with Chat is explore alt history scenarios.
I did this first with 4o. It was amusing to see how it created scenarios, but its responses were generally both superficial and unbelievable.
o3 was much better cause it found good sources and reasoned through them, but its prose sucked. And tables.
gpt-5 was only a small improvement over o3 here, with the q&a experience in the product the biggest difference.
But gpt-6 in chat is at another level with a better conversational style, improved reasoning and synthesis and of course visual intelligence. I must have spent 2 hours on my flight yesterday just going down different rabbitholes with it!
5
15
1,623
pranav retweeted
🗣️ The Agents API team has a shipping problem – we literally cannot stop. Here’s what we cooked up for you beautiful people this week:
💂 Auto Review: Just like in codex, auto-review will evaluate commands for potential bad behavior before executing.
🗜️ Async Compaction: Speed up your sessions by allowing compaction to happen in the background.
🕐 Time awareness: An experimental feature that allows the model to have a sense of time passing. This means the model will better act on queries like, ”timebox CI fixes to 10 minutes.”
🔨 Tool Updates: You can now update a session’s tools at any time.
💰 Better prompt-cache awareness: System prompts are now automatically cached across all sessions, saving you money and time