Ola. Ola.

What makes OpenDevBrowser different? - Supports Base64 (Encodes screenshot as binary‑to‑text ASCII characters like A–Z, a–z, 0–9, +, /.) - Supports CLI + CDP + Extension + UI annotation "npx opendevbrowser" "npm install -g opendevbrowser" To start: "npx opendevbrowser serve"

@browser_use Great tool. If you want agentic browser tool with extended functionality and OSS, try opendevbrowser.dev

Introducing: Browser Use CLI 2.0 🔥 The most efficient browser automation CLI tool > 2x the speed, half the cost > Easily connect to running Chrome > Uses direct CDP Try it now 🔗↓

@OpenAI "We're currently experiencing high demand, which may cause temporary errors". If you are running high stakes operations on GPT models and you are not on priority access, you may want to pause for a moment. Is this just one-off or something to expect going forward?

As AI agents and LLM adoption accelerates, lifestyle tasks e.g. shopping will become programmatic & need durable browser automation for agents to shop safely and reliably. opendevbrowser.dev deliver that. Let your agents shop & hunt for deals on major retailers in stealth

Opendevbrowser is solving an important problem for the next generation agent browser automation. The next release is going through release-gate stabilization. If you use AI agents in your workflow, try it. Fully OSS. Thank me later. opendevbrowser.dev

@ryancarson Have you tried opendevbrowser.dev ? It’s OSS and you will be pleasantly surprised.

Your homework for today ... Spend 8 hours completely optimizing your code factory and all of your agentic harnessing. It's tedious and frustrating, but it's worth it. I spent all of yesterday cleaning out automations, getting agent browser testing refined, culling skills and helper scripts. Really fine-tuned my code factory, and wow, it is worth it.

@HamelHusain I do this. After creating the plan with the agent, I open a new chat and ask the agent to "critically audit the plan, patch findings & repeat the audit-patch-audit-patch loop until plan is safe for implementation". I load custom built agent skills + tools in the process.

One thing that makes me feel that code factory has not arrived yet is the following experiment: 1.Ask a LLM to do an in-depth rigorous review of your code 2. In a new thread, as same/different LLM to consider those review comments independently and address issues it agrees with 3. Keep repeating until no new concerns I find that this loop always goes on for a ridiculously long time, which means that there is a problem with the notion of claude-take-the-wheel. This seems to happen no matter the harness or the specificity of the specs. It works fine for simple applications, but in the limit if the LLMs have this much cognitive dissonance you cannot trust it. Either this, or LLM are RLHFd to always find some kind of issue.

@Clark10x I have tried everything in that space with complex algos. I agree it’s very hard. I have combinatorial arbitrage here. Disable other strategies. Allow only FW baskets & projections. github.com/freshtechbro/o…

I've spent about 3 weeks and over $2,000 in Claude tokens setting up multiple agents to trade on Polymarket It was paper trading for a week and averaged about $80 per day It's been live for 2 days & completely broken All of these other people gotta lying on the timeline 😂😂

@Clark10x @victorleewyatt I have tried everything in that space with complex algos. I agree it’s very hard. I have combinatorial arbitrage here. Disable other strategies. Allow only FW baskets & projections. github.com/freshtechbro/o…

@victorleewyatt I’m gonna tighten it’s confidence and edge in order to place trades and see I am determined to make this work I’ve put in too much time already

@pvncher @RepoPrompt I totally agree. 5.4 is great and your instructions must be clear too. It takes instructions literally and follows them like a soldier. Harness, Context management & tools & clear instructions make 5.4 unbeatable. it cuts corners sometimes. So you need a verification tools

I’ve been using 5.4 non stop in the @RepoPrompt agent since launch with none of these issues. It uses codex app server under the hood, but the tools and system prompt are different. It’s still my absolute favorite model right now

@Dimillian I notice agents tend to cut corners with these rules sometimes. Add authoritative / deterministic validation with scripts. For each rule/guideline file, add a validation script to enforce compliance. Run the validation script to audit and surface non-compliance.

I've made quite a bit of improvements for my SwiftUI pattern agent skill, one of my favorite skill for working on SwiftUI apps. It distills all the patterns I use, with many examples, so the agent can do clean SwiftUI! github.com/Dimillian/Skil…

@reach_vb opendevbrowser.dev script-first for token efficiency comprehensive browser automation bi-directional frontend design canvas CLI / Headless / CDP / Chrome Extension

what did you build with codex this week? 👀

@MatthewBerman opendevbrowser.dev script-first for token efficiency comprehensive browser automation bi-directional design canvas CLI / Headless / CDP / Chrome Extension

What are you building this weekend?

@ryancarson Well Codex app does this already without linear. I have it running all day, all night on a plan and it’s getting everything right. At this point, getting the plan right is the most important step. Spend hours/days to plan, hand it off to Codex. Either 1 big chunk or piecemeal

This works well for me: 1. Plan mode in Codex on xHigh 2. Copy markdown plan 3. Paste md plan into Linear issue and Symphony takes it all the way to human review (or auto merge of low risk)

@robinebers github.com/freshtechbro/o…

@robinebers I recommend you try opendevbrowser.dev You will be pleasantly surprised. Does everything you mentioned and so much more. And it’s OSS. Just point your agent at the GitHub repo…& voila.

Codex' new computer use is genuinely insane I just asked it to use the open browser to check my last ten X bookmarks and it literally just did it If you have a Mac or a Linux machine... you can just run Codex to replace a bunch of expensive APIs no extra tools needed

@dhasandev @paper Pls do and share your feedback. It’s both a design canvas and browser automation tool. It’s script-first. So it’s as token efficient as possible. opendevbrowser.dev

@bishopdotun @paper this looks cool im gonan look into it

watch how Codex drives @paper to redesign a screen i wasn't satisfied with. modern product design is driven by two things: 1. taste (your general intuition on whats' good/bad trained by what you've seen) 2. iteration cycles paper can help you enable both. love this tool

@pvncher @RepoPrompt No. I haven’t. I will soon. I need time to play around with it first. Just the chat & context builder for now.

@bishopdotun @RepoPrompt Nice! Totally valid way to use it Have you tried using the agent directly?

@pvncher this is how I wired @RepoPrompt into the workflow. The agent builds context on the fly or after compaction and needs clarity. There is also a continuity ledger for alignment.

@Dimillian the clanker is working hard. Realtime bi-directional design canvas in your browser soon.