Some folks from @hackclub are visiting Sentry today and I was reminded of this post from Vighnesh. If you're looking for a program to financially support, I highly recommend considering them.

Links: [0] Handoff (No More Compaction): https://t.co/cYltAaveTV [1] Use Amp Free at Work: https://t.co/a22APAO4RT [3] 50% Faster Search Agent (w/Haiku): https://t.co/pFDOmGjmz1 [4] @mitchellh's Amp threads shared in https://t.co/5uHD3qaPaQ, and more like this soon.

More for less: the average paying Amp user now pays ~21% less per day vs. a few weeks ago, while # {threads,messages} {,per user}, users, & total revenue are all growing. We did this while keeping Amp's `smart` mode quality the same or better, and without subsidizing you (which

Grow faster on the world's largest group chat.

Need someone to join the Amp team and help us build ads infra & user flows for Amp Free. Backend, frontend, shipping many times per day, and working with customers. This role is San Francisco only. Lots of hard work and ambiguity. DM me.

Every bit represented in physical matter is a pbit anyway, even with ECC, if you account for random radiation and so on.

Met @GillVerd last year, chatted about this. Made intuitive sense. Maybe oversimplifying, but if computing something probabilistic anyway, you can rework it to use probabilistic primitives (pbits) if they conform to a known distribution. And pbits can become more efficient. GL!

The Left won't slow down so we won't either, someone's gotta make sense of the nonsense! Be sure to like and SHARE the stream!

You can now rename your threads in Amp on the web. This helps when you're sharing a thread to show your l33t agent skillz or to staple to a PR, and the auto-generated title isn't quite right. Very small. But...my voice is back after having lost it for the last ~week!

@compliantvc Have you seen this underground sham alcohol podcast by American tech founders? I don’t think they have a liquor license or a television broadcast license. Which EU form should I use to report it?

@gvrooyen It's at https://t.co/ig7C6ng12b.

Amp users are the best! @gvrooyen just published the senior eng guide to Amp he has been using internally. Self-recommending.

We removed slash commands from Amp. Instead we now have a command palette. It's better in every dimension. This frees up the prompt editor and sets us up for more powerful features and workflows.

The big Q for today is: with a more fragmented foundation model layer (more choices, but also some proprietary to their agent)…who wins and who loses? How does it change how people use coding agents? Do users primarily have affinity to the agent, or the model?

Big thanks to @gvrooyen and @sgjarmak for their help here in replicating these results!

[0] Next.js evals repo (fork with Amp support): https://t.co/D8Gg3boKBV [1] Amp search subagent evals:

Coding agent evals are useful for making your agent better, but not for proving it's the best (whatever that means). That's why we don't talk publicly about evals. But a lot of people ask, so here's a longer explanation of why they are not meaningful for agent-to-agent

Good example of how someone smart uses a coding agent

(page is https://t.co/NAGDN51PTr)

This is why a model selector doesn't make sense. 1. Make a tool-calling coding agent 2. Make the tools better 3. Some tools become subagents under the hood 4. Different subagents need different models 5. Maybe even their own models 6. So what would a model selector even select?

cc @Hesamation haven’t met but thanks for fighting the good fight

Coding agents work better if you use them right & adapt your codebase to the agent. Some devs get defensive if you say this, accusing you of blaming them & shirking responsibility. We make Amp for devs who want it to succeed and want to explore this crazy new world together.