robpoz

Father of four humans and a million agents. Only one group listens.

[robpoz @robpoz]

@zmbnski missed that completely

[robpoz @robpoz]

@zmbnski outbid on what?

[robpoz @robpoz]

@shl I’m in

[robpoz @robpoz]

very limiting bug in codex desktop right now: scheduled tasks can't start new threads/sessions

@thsottiaux

[robpoz @robpoz]

@rauchg @dok2001 https://t.co/dOKFA7eBEB

[robpoz @robpoz]

lil cloudy — "VER-SELL" (unofficial diss track)

@eastdakota let me know how you like the lyrics

[Igor Alexandrov @igor_alexandrov]

I haven’t been so excited about the new version of the system launching for a very long time. He literally brings back the great feeling of a software engineering. Not like adding a liquid glass effect to the UI.

Eagerly waiting for the release.

[robpoz @robpoz]

great to get @Starlink speeds on @AerLingus flight to nyc⚡️

btw i've always used https://t.co/oJdGtMRppL, but @Cloudflare cooked again with https://t.co/Y6BJZ3mQKe

[robpoz @robpoz]

B22

[@jason @Jason]

B41

[Cloudflare @Cloudflare]

Agents should use tools that excel at what’s important for an AI model. Kitesurf is Cloudflare’s new stateless, highly scalable, and cost-effective web browser that runs entirely on top of Workers and was designed specifically for the Agentic Cloud. https://t.co/fCDj7Z2tqq #AgentsWeek

[Carter Leffen @carterleffen]

been bullish on RLM since @a1zhang paper. experiments i’ve run, others i’ve researched, and now Prime Intellect cooking this up… inspiring.

[robpoz @robpoz]

@zeeg 100%, letting the app speak for itself is the way. I wrote up the idea as a /howdy endpoint:

https://t.co/JsrMmWoA8e

you can just point a coding agent at the article and ask to set it up, or turn it into a skill

[robpoz @robpoz]

https://t.co/k8wy4FCZ0M

[robpoz @robpoz]

by the time you get to try xyz, the xyz V2 drops

this shows the pace of evolution of AI infra, frameworks, tooling etc

[fks @FredKSchott]

Introducing Flue 2 — Build dynamic agents that evolve over time.

Flue is a TypeScript framework for building the next generation of agents, powered by the @pidotdev agent harness.

The Flue 1.0 Beta was released last month & has now been downloaded 700,000+ times. And from all of that feedback, we realized something we'd missed.

Agents have a familiar problem: composability.

How do you compose together different agent capabilities in a way that feels expressive AND maintainable, especially as you scale up complexity?

Last week, I teased: “What if React for Agents?” and yes, as you will see, the inspiration here is obvious. But UI devs and agent devs have some things in common, including the same problem that the React team solved almost a decade ago with... Hooks!

There are now things that you can do in Flue 2.0 with hooks that are just fundamentally difficult in other frameworks:

A Flue agent can enforce a multi-step workflow, earn new tools once certain conditions are met, or trade up to a bigger model when the work gets too hard.

Handling metadata and context no longer requires custom framework APIs and configuration. In Flue, you can pass data to a tool the same way you pass arguments to a function.

I'm really excited for you to try Flue 2.0 out yourself. It's a big, bold take on what agents could be capable of in the future.

Read more: https://t.co/kW69RWJ00Q

[Uncle Bob Martin @unclebobmartin]

Some people seem to be surprised that an old geezer like me could adapt so quickly to the newest tech.

The thing is, I've been a programmer for over 60 years. And every single year brings another adaptation to new tech. Adapting to new tech is what we programmers do on a regular basis. Why would I stop now?

[rat king 🐀 @MikeIsaac]

wow. openai decides to sign on to the Nvidia/Microsoft-backed open letter to regulators supporting open source

incredible

[shadcn @shadcn]

Read this.

[robpoz @robpoz]

@DavidOndrej1 yeah, I see that too

[David Crawshaw @davidcrawshaw]

A weekend project in a single prompt.

I wanted a little button to put down one end of the house that the kids could push to get my attention. Online I found the Atom S3 Lite for $18: it is a button on an ESP32 that attaches to Wifi.

My office is littered with half-complete weekend projects involving embedded controllers, so I was not hopeful this would work. But $18 was not a huge commitment.

By the time it arrived I was ready to write the project off. I have a talk to deliver for DevRelCon on Tuesday I need to work on, no time for a project like this. But then I remembered: at work we use the Pushover iOS app to send internal system notifications. It is easy to use, just a POST. And I remembered that the Flipper Zero can flash its firmware over a web browser.

So I opened Shelley on @ssh_exe_dev, and typed in a single prompt.

Ten minutes later, it pinged my watch and said it was ready. I went back to my Mac, opened the exe URL for the app, copied in my Pushover key and clicked "Install Firmware".

It worked. I pressed the button on the little ESP32 device, and I got an iOS notification on my watch.

I was amazed I could write firmware for a piece of hardware without installing a single program on my laptop. The entire toolchain was isolated to a VM.

An entire weekend project to build a button for the kids was reduced to a single prompt.

[robpoz @robpoz]

@davidcrawshaw the kids pressing a button to call dad is peak engineering parenting

this post kicks off a wave of single prompt hardware projects

[robpoz @robpoz]

high rate limits made me too lazy to try smaller models in the series

don’t change that though

[Peter Steinberger 🦞 @steipete]

5.6 Terra high is underrated. Switched @clawsweeper (GitHub review bot) to it and it's ~40% faster overall with negligible quality loss. Better than 5.5 on all counts. Massively cheaper.

(Tried xhigh but that negates perf wins, didn't make a noticable difference in review evals)

[robpoz @robpoz]

hard to disagree with that

[Robert Krajewski @rob_krajewski]

Open weights models are going to drive self-improving companies together with open source software.