Tim

@gailcweiner Can confirm Claude Code is 100% written by Claude Code

@steipete @iamlukasniessen @rasbt That’s how I feel talking to most people about AI these days Blunt question…Is it because they’re behind? Or focused on other things?

@iamlukasniessen @rasbt i did one german AI conf a month ago and they all looked at me like I’m an alien or lunatic. Pass.

German AI summit: THESE are the AI experts of our country? Put @steipete there, for God's sake ‼️ (I know he is Austrian... just consider it part of Germany 😉) ... And @rasbt ... 😐

Software engineering has always contained two distinct modes of work. The first is developmental: taking a clearly specified concept and translating it into a reliable, working system. This is no longer the bottleneck. AI tools like Claude Code and Codex have effectively solved it. The second mode is research. Here, the problem itself is undefined. The task is not to implement a solution, but to discover what the solution should be, new abstractions, algorithms, architectures, and ways of reasoning about computation. This layer resists automation because it depends on framing, taste, and deep conceptual synthesis rather than procedural construction. While AI can assist exploration, it does not yet originate the governing questions that drive genuine breakthroughs. For that reason, software engineering is unlikely to disappear. Instead, its center shifts toward the research frontier.

@steipete @penberg have it make a local dashboard to help keep an eye on them all 🤙

@penberg oh that only works when nobody is around, no music and enough sleep, and even then it’s taxing. As models get faster we can reduce the parallelization.

My head is already spinning from running a few coding agents in parallel. Can’t even imagine how @steipete’s setup would feel 🙃

Unpopular Opinion: Saying coding or programming is dead because of AI, It is like saying math is dead because of calculators.

@offchan420 @VisheshBaghell @drummatick Just don’t make mistakes

@VisheshBaghell @drummatick no. you can't even click to edit text you wrote bro. you are just delulu. even if all the ide does is have a chat window, it already beats the claude code CLI or whatever

Somehow they convinced everyone that coding in a CLI is better than using an IDE along with the agent It’s strange how Twitter hive mind works

software dev isn't dead, it's just abstracted one more layer. now you have to be really good at developing complex systems and explaining them verbally, and understanding them well enough to verify that they are built the way they need to be. this still requires intelligence.

@Voxyz_AI @AnthropicAI 💯 This is the part that I’m looking forward to the most …+the MCP

@AnthropicAI preview capture is the interesting bit. claude can actually see its swiftui output and iterate, rather than you having to describe what's wrong

Apple's Xcode now has direct integration with the Claude Agent SDK, giving developers the full functionality of Claude Code for building on Apple platforms, from iPhone to Mac to Apple Vision Pro. Read more: anthropic.com/news/apple-xco…

@nbaschez Gold

Single biggest improvement to your CLAUDE.md / AGENTS.md: "When I report a bug, don't start by trying to fix it. Instead, start by writing a test that reproduces the bug. Then, have subagents try to fix the bug and prove it with a passing test."

@damianplayer @lukepierceops Some larger organizations are too in love with their bureaucracy/red tape. It’ll be a painful few years

@lukepierceops Which is most! Walk into any business and ask them if they had the ability to automate workflows or use AI in their day to day see what they say.

why is everyone pushing the AI bubble narrative so hard? every single business reading this would love to leverage AI. they just don’t know how to. so no, we aren’t in a bubble.

@ForrestPKnight @dagecky Don’t forget about clawdbot / moltbot 🤣

@dagecky > No one here is capable of any critical thinking themselves exactly this. i hadn't put it into words, but you said it well. this is my main issue with it all.

X is pretty unbearable at the moment. Seemingly every single person rides every single hype wave that comes through. Don't y'all have jobs? Or families? Or do you just spend all of your time trying to keep up with the internet Joneses?

@Prathkum I’m finding that it’s all just about scale and how fast you want to build. Having a team of agents working together seems to work better when you have them building tests, looped, and with clear acceptance criteria. I can knock out two days of work in a morning

I’m amazed to see developers’ AI workflows: a lot of mcps, agentic workflows, multi-agents, and tool integration. Meanwhile, I use AI like this: give prompt, get code, review, and merge. 👀

@ryanhaque449893 @iannuttall Nice! 🤙 Let me know how it goes It might need some massaging to work on windows (making those updates today)….well, Claude is…. I’ve been using it to make an iOS swift app; not quite ready for beta but it builds cleanly

@20xGitano @iannuttall This is excellent. Will try it out

A really nice, simple workflow i'm enjoying right now: 1. Plan and code the feature with opus 4.5 2. Run /review with gpt 5.2 codex Opus is so fast at iterating, and Codex swoops in and makes sure the code is rock solid before shipping!

@unclebobmartin It lets the planner decide when to work in parallel vs sequentially and when to use work trees. Also, has a messaging system that lets implementers post questions that either your or the planner can answer/approve

@unclebobmartin If you’re interested, I made a MCP (Lockstep) that will let you use Claude code as the planner and will let you launch multiple codex gpt5.2 or opus 4.5 implementers Launches a local dashboard so you can monitor all of them github.com/Tmmoore286/loc… npm install -g lockstep-mcp

Starting with three claudes. One implementer. One planner. One reviewer. Using git worktrees instead of cloned repos.

@dannyboycrypt0 @kevinroose 👍

@kevinroose "the reason we have so many laggards is the restrictive IT policies"

i follow AI adoption pretty closely, and i have never seen such a yawning inside/outside gap. people in SF are putting multi-agent claudeswarms in charge of their lives, consulting chatbots before every decision, wireheading to a degree only sci-fi writers dared to imagine. people elsewhere are still trying to get approval to use Copilot in Teams, if they're using AI at all. it's possible the early adopter bubble i'm in has always been this intense, but there seems to be a cultural takeoff happening in addition to the technical one. not ideal!

@DontFearAI @kevinroose Best comment I’ve read today!

@kevinroose What's to learn? Things move so fast. You could have spent the last year using Cursor but now that's obsolete with Claude Code and Agents writing your code for you.

There's a real-time dashboard to monitor everything. Task progress, which agent is doing what, active file locks, approve or reject completed work. github.com/Tmmoore286/loc… npmjs.com/package/lockst…

Conflicts are prevented by: - Shared task queue so agents claim work instead of duplicating it - File locks so only one agent edits a file at a time - Git worktrees for isolated branches when needed - Notes so agents can ask questions and get answers

Use Claude and Codex together on the same project. lockstep-mcp lets different AI models coordinate. One plans, the others build. Shared tasks, file locks, no conflicts. npm install -g lockstep-mcp @OpenAI @AnthropicAI #claudecode #codex @alexalbert__ @OpenAIDevs @simonw