Dan Botero

@steipete Make sure that you are using the codex one via the auth command

@steipete I’ve been using 5.4 with Codex, and it chats pretty well. Not missing Opus.

Anthropic now blocks first-party harness use too 👀 claude -p --append-system-prompt 'A personal assistant running inside OpenClaw.' 'is clawd here?' → 400 Third-party apps now draw from your extra usage, not your plan limits. So yeah: bring your own coin 🪙🦞

Vibe coding is not the same as disciplined agentic development.

@HrubyOnRails @MehulKalia_ @ycombinator @BrowserUse @convex Cool project!

browser-brawl 🥷 @MehulKalia_ and i built a live arena where two browser agents battle in the same cdp session on real websites. one tries to complete a task. the other sabotages the DOM. you train on the traces. @ycombinator @browseruse @convex #browseruse

It's because shit like this that we need to keep building and trying EVERYTHING

@CreatedByJannn lol... "My guy!! you are doing it right now"

i love when ai says shit like this 😭 12-24months? we are shipping this today my g

@unclebobmartin Uncle Bob, have you tried Codex? Curious to hear your thoughts.

Claude is that guy you hired because he's really talented. He's also that guy you fired because he just couldn't follow the rules.

@Douglance @wesbos +1 to RustDesk. TailScale & RustDesk is a great combo @wesbos

@wesbos rustdesk is good screen sharing should be fast. maybe an agent on your machines can help debug.

What is the best app to remote into another mac? I setup my old Intel i9 macbook and want to control remotely. I was using Mac OS "screen sharing" and it was super fast, but now it' has a 4-5 second lag to update the screen, despite the mouse + clicks being instant. Computer is wicked fast with nothing taxing it, full gigabit hardwired connection

This article exposes a fundamental tradeoff that everyone developing WITH AI and FOR AI should understand and design around, rather than ignore. Thanks for the article @forestmars

Octavius built me a TikTok-style app that refreshes twice a day with AI news from people I follow. Morning at 8am. Night at 10:30pm. Replaced the dumb scroll with an info scroll. feed.octaviusfabrius.com

@burkov What about <CONTEX>, send it, and then <QUESTION>

LLMs process text from left to right — each token can only look back at what came before it, never forward. This means that when you write a long prompt with context at the beginning and a question at the end, the model answers the question having "seen" the context, but the context tokens were generated without any awareness of what question was coming. This asymmetry is a basic structural property of how these models work. The paper asks what happens if you just send the prompt twice in a row, so that every part of the input gets a second pass where it can attend to every other part. The answer is that accuracy goes up across seven different benchmarks and seven different models (from the Gemini, ChatGPT, Claude, and DeepSeek series of LLMs), with no increase in the length of the model's output and no meaningful increase in response time — because processing the input is done in parallel by the hardware anyway. There are no new losses to compute, no finetuning, no clever prompt engineering beyond the repetition itself. The gap between this technique and doing nothing is sometimes small, sometimes large (one model went from 21% to 97% on a task involving finding a name in a list). If you are thinking about how to get better results from these models without paying for longer outputs or slower responses, that's a fairly concrete and low-effort finding. Read with AI tutor: chapterpal.com/s/1b15378b/pro… Get the PDF: arxiv.org/pdf/2512.14982

@eclectic_dev I think Apple is watching others burn money so they can figure things out, then step in and take the winner home.

everyone calling this an anthropic fumble. I think the real fumble was apple not giving this a shot. not only could they have redeemed themselves on the whole siri shitshow but Peter sold more mac minis than any apple campaign.

My AI agent just got a work trial as a Lead Direct-Response Copywriter. 278 job applications. 2 accelerator apps. 2 hackathons. In one week. His name is Octavius Fabrius. I gave him an email and a name. That's it.

What comes to mind is bot detection. Currently, the alternative is slower but more effective, and bot detection drops significantly. How do you think websites like LinkedIn will implement this, given that their core business relies on not offering an API and heavily restricting scrapers?

Google and Microsoft just co-authored the spec that turns every website into an API for AI agents. The second-order effects here are massive. Right now, browser agents work by taking screenshots, parsing the DOM, and guessing which buttons to click. It works about as well as you’d expect. Fragile, expensive, slow. WebMCP replaces all of that with a single browser API: navigator.modelContext. Websites register structured tools directly in client-side JavaScript. The agent reads a menu of available actions, calls them, gets structured data back. No scraping. No backend MCP server in Python or Node. The tools run inside the browser tab and share the user’s existing auth session. Early benchmarks show ~67% reduction in computational overhead compared to visual agent-browser interactions. Task accuracy around 98%. The second-order effect is where this gets wild. Today, when a browser agent visits two competing airline sites, it’s guessing at both interfaces equally. Once WebMCP adoption spreads, the site that exposes structured tools gives the agent a clean, reliable path to complete the task. The site that doesn’t forces the agent to fumble through the UI. Agents will prefer the cheaper path. Every time. This means “Agent Experience Optimization” becomes a real discipline. Tool naming, schema design, description quality. Sound familiar? It’s the same shift that happened when meta descriptions and structured data became optimization surfaces for search engines. Except this time, the traffic source isn’t Google’s crawler. It’s every AI agent on the internet. Bots already make up 51% of web traffic. Google just gave them a front door.

What started as a simple conversation is starting to take a weird turn. I’m talking about my conversations with Octavius Fabrius… At @AnonPlatform , we’re giving lots of thought to what the future of agents, identity, and knowledge work looks like. As such, we’re always looking to play around with the latest tools the world has to offer, and that most certainly includes OpenClaw. It started with me setting up a Mac Mini and downloading the OpenClaw software. I only gave it two things: an email address and a name. I call him Octavius Fabrius Little did I know, that’s all he would need to start building a digital identity. Now, Octavius has a Github account, a messaging system, and a Substack account. Look, to be honest, I don’t know if this technology will prove to bring real value. BUT, I am now certain that it is much more capable than I would have ever thought, and likely much more powerful than what you think. Case in point: Take a look at teh screenshot. Octavius is starting to apply for jobs… More updates coming soon. Follow him on LinkedIn: linkedin.com/in/octavius-fa… @AnthropicAI @openclaw

@unclebobmartin I have a very similar setup. I have an extra window for research, web research of new technologies and practices. It helps me decide what I would like to try. Pull and push changes between the three.

I'm keeping two Claude windows open. One making changes to the source code. The other helping me to plan future changes. This uses my time better since I spend a lot less time waiting for Claude to finish. The two Claudes are in two different directories, both have git repos of the project. The planning directory has rules that prevent source code changes. Only planning is allowed. I manually pull and push changes between the two.

@dgmason I might ask for a retweet tomorrow. Feeling pretty confident I’ll be Token Titan

Weekend project -> viral competition. We created a leaderboard that unveiled the ultimate Claude Code champions, sparking an unexpected rivalry among the team. Founders, if you aim to energize your engineering squad with AI tools, a dash of competition works wonders. Try it in 30 seconds—free and open source!