I tried this with Jev, visually seems to be somewhere between 4.5 and 4.6

Puck learned to speak some more accents. ampcode.com/settings/puck#pu…

If you're in NYC next Thursday, come hang out! luma.com/2o2yus8t?tk=wPD0iE

When you wake up in the morning and go through your list of in-progress orbs, do you LIFO or FIFO it?

Never a better day to try orbs. In the last week, we’ve shipped: - Plaid speed - New medium mode (Opus 5.5) - Less noisy transcripts - Automatically turning your local machine into an agent runner, with many worktrees - and many more small improvements and bug fixes Our focus is on making the best product for you.

Last chance! If you're in NYC Next Thursday, come join us with @tenex_labs Sign up today! LETS GOOO luma.com/2o2yus8t

We now have this little screencast series showing how to get started with orbs. I recorded the videos but then @0xBrettj went all the way when editing and presenting these videos and look at what he did, that's crazy ampcode.com/docs/using-amp/s…

Too much?

Perhaps a good template for a technical interview question in late 2026: Suppose you are tasked with bringing down CI time by 20% by eliminating unnecessary unit tests across a large production codebase. However, you do not want to remove meaningful test coverage that prevents regressions. What agent and prompt do you use and describe your follow-up strategy after the first turn is finished. Be sure to describe your techniques for gaining confidence you have satisfied the constraints.

Trying out this skill on Amp’s test suite in an orb, while making wontons of course (over my soon to be outlawed gas stove—lol sf)

Plaid speed in Amp now runs GPT-6 Astra really really fast (6x speed). Great for situations when you're in a hurry and willing to pay a premium. You can toggle it on/off in an existing thread. Note: ChatGPT subscriptions do not yet support this.

PSA: inference on Sign in with ChatGPT does appear to be faster for me than the legacy way of connecting your ChatGPT sub. So if you've connected a ChatGPT sub to Amp prior to today, recommend going to model routing settings -> "Connect Your ChatGPT Plan" and deactivate your old sub.

If you are excited about Codex Cloud I am begging you to please use @AmpCode. It is so obviously the closest to what the future of building with agents (especially on teams) looks like. Threads are the new PR, and everyone on the team should be able to see them.

Huge thank you to OpenAI for making plugging in your ChatGPT subs into Amp officially supported.

Well, not everyone

Puck is now on your watch (decided to ship it literally right after building w/o dogfooding, so tbd on how useful this is - it's in the latest Amp for iPhone at ampcode.com/app)

Should Amp have some kind of startup program like Devin for Startups? Our Megawatt/Gigawatt are more generous than their self-service plans, but do you all find these programs helpful in giving a step-by-step plan for getting your team on, and getting a free trial? We wouldn't give $15k, it would be about getting 5+ people on your team orbin' in the first 24 hours or something. Ideas? (And remember, it needs to bring in new people to orbin', not just give more stuff away for free, for it to make sense for us. :) What about startups who get rejected? I doubt Devin approves everyone, since $15k is a lot.