J. Yeomans

Built a tool to bridge AI agents and your web UI. Click any element to get CSS selectors and identifiers exported as markdown. Framework-agnostic, with built-in adapters for Angular, React, Vue, Svelte, or vanilla JS.

@dawiddrzala @GeorgeJourneys It’s legit: x.com/ffmpeg/status/…

@GeorgeJourneys That's why this story dose not make sense, who verified that vulnerabilitis another AI ? Mayby they are hallucinations. Till any of the software teams no going to confirm what was the bug and fix , this is just bla bla

So, basically, if Anthropic was not a US company, we’d be facing zero days with multiple unknown points of attack on virtually all of our systems to an adversary who developed this capacity before us.

@andrewfarah I bookmarked this

sharing my first open source project a CLI for downloading and syncing your X bookmarks locally so your agent can access them. it's free › npm install -g fieldtheory › login to your X account in a chrome tab › ft sync (done!) bonus: › ft viz › ft classify

@kr0der @towheretobegin You don’t need to read the plan. The plan is there to keep the agent on track during implementation vs it having to figure out implementation across dozens of user messages. Easier to one-shot more complex changes by having a plan for the agent with validated assumptions

@towheretobegin kinda, it's more like it's hard to read the giant wall of text plan at the end (rip attention span)

slowly starting to use plan mode a LOT less nowadays i realised whenever i use plan mode, it generates a gigantic plan and then i dont read it and hit build out of laziness having a meaningful conversation with the AI agent to discuss implementation feels a lot easier 🤔

@doodlestein That was my thoughts 👍

@YeomansIII I don't see a point to it, just seems slower and will use more tokens.

Agent Coding Life Hack: This is like the coding agent equivalent of shining a blacklight on your clean-looking black shirt and seeing just how filthy it really is: ❯ First read ALL of the AGENTS.md file and README.md file super carefully and understand ALL of both! Then use your code investigation agent mode to fully understand the code and technical architecture and purpose of the project. THEN: I need you to carefully and completely look across the ENTIRE project for *anything* that is a hard-coded constant in the code which really should be dynamic in order to be correct and robust. Also look carefully for any "TODO" or the words "will" or "would" in a comment, indicative of unfinished code. --- I was truly horrified by how much stuff this turned up. If you are, too, then try following it up with this one: ❯ OK, please fix absolutely ALL of that now. Keep a super detailed, granular, and complete TODO list of all items so you don't lose track of anything and remember to complete all the tasks and sub-tasks you identified or which you think of during the course of your work on these items!

@mckaywrigley Down to try, hmu

looking for a handful of people to test something new... i've been using it for a few months and am prepping to share. if you're a fan of claude cowork, openclaw, manus, perplexity computer, etc then you're a perfect fit. this will self destruct in 4hrs - please dm or reply.

@yiliush @_Sagiquarius_ This app is honestly exactly what I was looking for. Thanks! Trying it out today.

@_Sagiquarius_ Good to be sure about stuff

source code now available github.com/collaborator-a…

@LDiken @petergyang @trq212 This got better last week but still happens.

@petergyang you can toggle it within /config but the session will time out after a while, then you will have to go into each of your open sessions and disable and re-enable /remote control. @trq212 could we have remote control not timeout/disconnect after a while?

If I have a bunch of local projects in Claude Code do I need to turn on remote control for each one to be able to prompt it from my phone? Wish it would just work instead as long as I have my computer on.

@waim @noahzweben Yes, it calls the CreateCron tool

@noahzweben Does the agent ever call loop on its own? I have asked Claude to write skills to wait in stuff and I’ve never seen it self author a loop

Loops now run for up to 7 days instead of 3. Let me know what you’re /looping on!

@bobsanch @trq212 @maxbittker Same here. I get worried they are pushing out so many new features the “old” ones from 2 weeks ago get forgotten about so quickly. They should assign a long-running agent to each feature that takes ownership of maintenance and fixes.

@trq212 @maxbittker OH SHIT. Btw just wanted to make sure remote-control's buggy-ness is on your radar (lots of crashes/ghost prompts). Or if it's just me lmk what you'd like me to send your way

Claude code would be a lot more empowering if you could have a side-conversation while it works. When I’m doing hard debugging tasks it’s hard to balance letting it cook with developing my own mental models and steering

@raymondjacobson @atmoio @theo I agree with this. And you can still write garbage “code” into the AI compiler.

@atmoio @theo maybe more deterministic, but you can still can easily make mistakes programming in C and the compiler can also do strange things and cause bugs. AI is much more complicated than C but it’s essentially just acts as a compiler I think. what’s your definition of programming?

I was a 10x engineer. Now I'm useless.

@TurnUpTheVix @MoonlitPython @DJSnM Facts.

First pass it came back with the right values? What happens when you ask it to do a performance tune of the code? The shit works if you just ask for what you want instead of hoping to get production level code out of basic requests with no second pass. "Industrialized woodworking will never work because you have to sand it by hand" right before they show you the sanding step that's also automated.

When I was first forced to switch from hand tuned assembly to C I complained that the compilers were generating slower code. But the truth is the expanded accessibility that compilers offered was far more important than my loss of performance. And, compilers got better.

@trq212 Will it respond with voice soon?

Voice mode is rolling out now in Claude Code. It’s live for ~5% of users today, and will be ramping through the coming weeks. You'll see a note on the welcome screen once you have access. /voice to toggle it on!

@roman01la @jloganolson Iterate

@jloganolson That’s a bad design, usually you’d smooth the corners for improved durability I guess Gemini just smashed together a bunch of primitives without thinking through potential durability issues

Very impressed with gemini pro one-shotting an enclosure for this speaker based on a few photos with calipers (even got the mounting holes aligned!)

@Yossi_Dahan_ @claudeai Writing the right prompt is still super important :)

@claudeai Claude: runs the numbers, writes the report Him: “Claude and I”

Introducing Cowork and plugin updates that help enterprises customize Claude for better collaboration with every team.

@PaulRoundy1 @RWMaloneMD I was gonna say, we still have very low rates comparatively…

@RWMaloneMD Virginia's power costs have risen substantially less than many other states that have fewer data centers. The datacenters can't be the main driver of high power costs.

This is a photo I snapped flying over N. Virginia recently. There are currently 685 data centers in Virginia, with many more being built, and a total of 427 Dollar General stores. Virginia has more data centers than any where else in the world. Somehow that ratio doesn't seem right...

@mrxseek @claudeai Yawn. Where is Opus 6???

@claudeai When is Sonnet 5.0 coming.???

This is Claude Sonnet 4.6: our most capable Sonnet model yet. It’s a full upgrade across coding, computer use, long-context reasoning, agent planning, knowledge work, and design. It also features a 1M token context window in beta.

@sketch @wolfr_2 Nice work! Maybe Johan is referring to MCP? Skills are the new hot thing and agentic models have proven to work better with CLI commands vs MCP calls - which take up a lot of context. Worth considering offering a Sketch CLI and skills with basically the same endpoints as the MCP

@wolfr_2 Go on Johan, enlighten us…

So far behind and they don’t even know it

@Erdayastronaut @WG_Mojo Tim, I think the amount of money in this Nvidia buck will/is creating massive opportunity for Nvidia competitors and startups making silicon and hardware. Nvidia is massive now because they had the hardware everyone needed when they needed it. Now everyone else can catch up.

@WG_Mojo I don't doubt the core technology of AI is in our lives from here on out, but when AI tech accounts for a third of the US market value... and with everyone passing that NVIDIA buck around, something feels bound to burst.

I just hope if (when?) the AI bubble bursts all the space computing is still of value. I suppose if you can make space compute cheaper than ground based, then it's a competitive advantage even if AI doesn't "pan out" so to speak.

@cybrtrkguy @DirtyTesLa Autonomy is: being able to put a cool wrap on my car to show off; being able to get up in the morning at my rural home and head to work with a full charge; leave my shit in the car while I’m working or at the store including car seats, dog blankets, umbrellas, junk, and trash

@DirtyTesLa Americans have largely turned against EVs because of the structured nature of ownership. Limited range and unfamiliar tech. I believe other countries will embrace autonomy at scale (like Norway has adopted EVs) but America will be autonomy's final frontier. Will take decades.

Is Tesla ownership coming to an end?