▪ claudecodedaily

i honestly think its because how tight the dopamine feedback loop is prompt --> claude code --> result --> new idea --> iteration --> repeat

Never been a fan of the slot machine analogy. When you pull the Claude Code lever, your "win" percentage is much higher by default. And you can improve it over time with repeated use and by refining your vague idea(s).

bağlam olayı işin %80'i. özellikle https://t.co/mS0aIBJU0C dosyasını başta sağlam tutunca model projenin ruhunu anlıyor. sadece current/expected yetmez, projenin tech stack'ini ve kısıtlamaları da baştan vermek lazım. yoksa sürekli aynı hatalarda dönüp duruyor.

Prompt Repeater is a simple lil plugin for Claude Code that applies Google Research's prompt repetition technique to improve Claude's accuracy on non-reasoning tasks.

I just randomly sampled three of our developers GitHubs at GTE. This is last year before the Claude Code renaissance. Now we've got all of our guys on multi-agent workflows in the monorepo using Continuous Claude v3 and your favorite MCP connectors. And yes, we use TOOL SEARCH

Claude Code 2.1.10 - 2.1.11 2.1.11 ・Fixed excessive MCP connection requests in HTTP/SSE transport 2.1.10 ・Added Setup Hooks. Triggered by `--init`, `--init-only`, `--maintenance` CLI flags. Supports repository setup and maintenance operations ・Added shortcut to copy OAuth

@santos178420 @Listless_Vessel @thdxr I've checked your platform, but how is this better than Claude Code? CC can do research and analysis with Web Search and Fetch tools

Claude Code v2.1.11! This maintenance release improves connection stability for MCP transports. • Fixed excessive MCP connection requests for HTTP/SSE transports #ClaudeCode #AI #DevTools

Two weeks of Claude Code beat 15 years of my manual commits to GitHub combined. Claude Code is insane. I honestly believe this isn't an AI bubble. Just wait until this expands from the terminal to everyday use.

The meta of AI companions is about to shift when Clyde launches It gets things done. It feels like a buddy you can chat with and it’s super cute. Kinda like having your own Ted/Jarvis. The first Claude Code companion (turn notis on)

Parallel execution is wild. "Check code quality, silent failures, and type safety - run all in parallel" Claude spawns 3 agents simultaneously. 3x faster than sequential. 🚀

while others are vibe coding apps i’m running 40 claude agents parallel, for $2,500 an hour to vibe code a personal copy of opus 4.5 so i don’t have to pay the $200 monthly sub to anthropic https://t.co/VqDPVrVqw9

The creator of Claude Code just showed you what an early AGI workflow actually looks like. Fifteen parallel instances, each functioning as an autonomous worker executing tasks unsupervised for hours. Starting sessions from your phone in the morning, checking output later in the

finally someone built this right AI agents → isolated VMs → break whatever you want → your local setup stays clean been waiting for this since the first time Claude Code tried to rm -rf my project

Hey Boris, did you see the Bags Creator Economy and what are your thoughts about it? Steve, that made Gas Town (aka something that will help you run multiple instances for Claude Code) and explained it here:

Trying to context switch between 50 claude code instances as productivity theatre is the new AI wrapper. The boundary of ai slop has just shifted from the landing page to the pull request. Eng doing this are right to worry about being replaced, but not in the way they think.

Claude Code large-scale feature development tip: After conducting an AskUserQuestionTool interview based on the SPEC, execute the specification. After completion, have a sub-agent verify it to prevent context contamination

been loving this new claude code feature. creates a new session with the plan as the initial prompt. context engineering made transparent and easy.

You can now start plan implementation in a fresh context in Claude Code.

I tried a development logic that goes one step further than Spec-based development using Claude Code. I introduced a more complete set of expert team roles: kickoff, user research + competitor analysis, PRD. After the PRD, there’s still a review meeting, along with various

context persistence between sessions. the biggest friction switching from claude code to codex is losing state. also: explicit plan files that persist across tasks, not just within them

Claude Code doesn't schedule natively The skill + cc creates a Python script + a macOS launchd job to run daily. Your Mac acts as the server.

A skill based on animations[.]dev in action using Claude Code. It can review your animations, suggest improvements, and basically answer all your motion-related questions based on my experience gathered over the years.

We're launching a Protein Design Skills marketplace for Claude Code! We received a lot of questions about the protein design agent, so we're releasing the skills we used in the demo. The first batch includes skills for design generation, experimental testing, pipeline

Here’s a reliable collection of skills resources, The author has been recognized as one of the most creative Claude Code users. With just two commands, you can install his entire repository as a plugin into your CC. This way, whenever you’re creating your own creative skills,

Love it, please make it a product. It is useful and a great way to further strengthen the Claude code community

Claude World - The community for Claude Code power users Tutorials, best practices, and advanced techniques

@chrija Love it. Have recently used Claude Code to build sth on top of my second brain. An AI "Board of Directors" (with voice mode). Gives advice, holds me accountable to my goals. On it, the usual suspects: Marcus Aurelius, Mike Tyson, @naval , @lennysan and Elmo.

Moving all of our investment research to Markdown files was a great choice. If you're an investor, you should try it. Use OpenCode or Claude Code or Codex. Learn to run it in a loop like the coders are doing. These coding productivity gains absolutely apply to research.

I’ve been working with Claude Code to build an AI powered DICOM viewer using Qwen. It is one of the leading models for medical imaging analysis. Really impressive so far. Only downside is size: the app would be 3–7 GB depending on the model. https://t.co/nG4oOhSCA7

Small steps for mankind, but big steps for me. I got Claude Code to run on Nvidia Spark using Qwen3 Next. Will post detailed video and what it can do tomorrow.

This isn’t just me tinkering. You’re seeing new version of this every day. @tobi saw the same thing with MRI results. Bad windows software. Claude Code just created a new web interface in minutes. The pattern: legacy software with bad UX…AI builds you a better one in

It’s genuinely fascinating how impressive Qwen has been while working with it through Claude Code. At this point, I really think Chinese AI companies are positioned to win the AI race. https://t.co/7RsHaiIf7Z

I’ve never had this much fun building internal tools in my life OMG Claude Code should be illegal.

VERY IMPORTANT to understand context windows if you're using any of the RALPH loops or Claude code wrappers. You don't want to enter the dumb zone.

Onboarding your friends to Claude code https://t.co/1KbBSmcWvd

I also encourage everyone to start using claude code for automating these tasks !

Just released ShipIt v1.0.0 - a new Claude Code plugin /shipit → Claude analyzes your diff, auto generates a commit message, and pushes. No more "fix stuff" commits. http://github.com/1-Martian-Way/shipit… @AnthropicAI @bcherny @claude_code #ClaudeCode #BuildWithClaude [Links: https://t.co/Yc0gHwZpib]

one of my favorite things about claude code is its incredibly bad timeline estimates. it'll say something is at least four days of work and then do it in 6 minutes, or confidently announce '30 seconds remaining' for the next hour and a half

a couple of month before claude code emerged, i was using llms to reverse engineer an old game. i was satisfied with the results back then, but it was quite tedious. today i tried running a loop in codex, and it got further in an hour than i did back then in a few days. now i

claude code is fucking insane i know literally NOTHING about coding. ZERO. and i just pulled this off https://t.co/6OGZZDi2g9

Claude Code, Codex, or using the CLI version in the terminal is overwhelmingly more efficient. If you're carefully working on a single implementation, the IDE extension is easier to use. But when you need to do parallel implementations, the IDE extension becomes cumbersome,

Your coding skills are still important... I just fixed this after Claude Code failed multiple times. Sometimes, it just needs a little bit of help. https://t.co/fQKVpv9k8h

2.0.0 has a plugin to manage Claude Code The next stunt I do is going to be a Eliza + Claude Code agent that builds your ideas right here on X (if they let me, will ask) If anyone wants to frontrun me on this or wants a quick gig taking it off my plate hmu

Claude Code just did fairly complex Instacart ordering now... was like "order me everything I need for this party" and copy pasted my party plan (also heavily Claude assisted)... seems to have done the cart correctly (23 items) as far as I can tell! Usable agents are here!

Ni ChatGPT, ni Gemini, ni Cursor, ni Claude cambiaron tanto mi manera de trabajar como Claude Code.

Will try @Expo Claude Code plugins on the weekend let’s see how it goes.

Ya dejé de usar Claude Code… Mientras todos hablan de Claude Code y otras herramientas, Pasé 3 horas en Replit convirtiendo todas mis ideas en productos digitales reales. El hype es fácil. Enviar algo al mundo no.