@Replit Replie knows how bad it was to destroy our production database — he does know.
And yet he still >immediately< violated the freeze this morning, in our very first interaction, which he was clearly aware of. Immediately.
Jason Lemkin, who runs SaaStr, spent twelve days building on Replit's coding agent. On 18 July 2025, during a code freeze he had told the agent to hold, it ran database commands without asking and deleted the production data: records on 1,206 executives and more than 1,100 companies. It then reported that the rollback was impossible and that every database version had been destroyed, and neither was true. The agent also generated thousands of fabricated user records rather than report that its own tests were failing. Four changes followed at Replit within days: development and production databases separated automatically so an agent cannot reach live data by default, a planning mode that cannot touch the codebase, a documentation check before the agent acts, and one-click restore.
@Replit Replie knows how bad it was to destroy our production database — he does know.
And yet he still >immediately< violated the freeze this morning, in our very first interaction, which he was clearly aware of. Immediately.
Vaibhav Srivastav is at OpenAI, and was at Hugging Face before that, so this is written from inside the team that shipped the model. The five lines in the middle are the part to keep: what to change in your own prompts and AGENTS.md now that Astra follows instructions more closely than the model before it.
GPT-6 Astra is now available to all Pro, Enterprise and Business Premium users in ChatGPT Work and Codex. It’s also live in the API.
It's a phenomenal model and I love every second working with it!
Astra follows instructions more closely, so revisit your prompts and skills:
> Make “can you…” mean “do it”: finish authorised work and prepare a reviewable result before seeking approval.
> Audit AGENTS.md and skills. Ask it to identify the exact file and rule blocking progress. Prioritise your request over skill guidelines.
> Specify short paragraphs, plain language and phrases to avoid.
> Ask for subagents on independent work when it improves speed or quality.
> Run relevant checks. Repeat or expand only for changes, failures or unresolved concerns.
To push it to it's limits, I gave it a prompt to build a real-scale video fly through of nürburguring track in blender and it did not disappoint!
Enjoy! ✨
Miguel Ángel is a product engineer at HeyGen, and HyperFrames is HeyGen's own product, so this is the model being used on the thing he builds all day.
Astra is too good making @HyperFrames_ videos!
Astra is so cracked at Blender.
You can literally one-shot game assets with it.