Thomas Witt

$285B wiped out because Anthropic posted a text file on GitHub.

Not a new model.Just ~2,500 lines describing how to review contracts—first-year law school stuff.

"Sophisticated" traders panic-sold on a headline without checking anything.

Apparently traders have no idea what they trade on.

I just released the aws-sdk-http-async gem: github.com/thomaswitt/a... - an Async HTTP handler plugin for the AWS SDK for Ruby, built on async-http

More information on the background: thomas-witt.com/blog/aws-sdk...

#ruby #rails #aws #async #fibers

Finally codex supports conversation steering as experimental feature and is catching up to claude code.

Also worth reading regarding codex as agent: developers.openai.com/codex/g...

Tech predictions for 2026 and beyond We’ve caught glimpses of a future that values autonomy, empathy, and individual expertise. Where interdisciplinary cooperation influences discovery and creation at an unrelenting pace. In the coming y...

Werner Vogels: “Read the headlines telling us that AI will make developers obsolete. In the 2000s, operations engineers expressed similar concerns when cloud computing arrived. Instead, it created an explosion of new companies, and new engineering roles”
www.allthingsdistributed.com/2025/11/tech...

Some examples of people quickly accomplishing ambitious things together: patrickcollison.com/fast

Manage Claude's memory - Claude Code Docs Learn how to manage Claude Code's memory across sessions with different memory locations and best practices.

The latest 2.0.64 release of Claude code is interesting. Besides finally named sessions and async agents, they support so-called rules. Maybe that's finally a solution that claude follows coding styles better:

code.claude.com/docs/en/memor...

(https://github.com/anthropics/claude-code/blob/main/CH

Writing a good CLAUDE.md `CLAUDE.md` is a high-leverage configuration point for Claude Code. Learning how to write a good `CLAUDE.md` (or `AGENTS.md`) is a key skill for agent-enabled software engineering.

www.humanlayer.dev/blog/writi...

First tests with OpenAI's new gpt-5.1-codex-max - even on xhigh - are very disappointing. It doesnt follow instructions well and I can see nowhere that its working 24 hours on tasks - it basically gives immediately answers to very complex tasks, which can't be right.

Coding Agents Tip: When implementinglarge features, you MUST write a spec doc first, divided in phases with not too many subtasks (#Claude in ultrathink plan mode isquite good at this) as a md file. Then ask #codex to implement it and stop at each phase. Then run claude and another codex to review.

I finally CRACKED Claude Agent Skills (Breakdown For Engineers) Claude Agent Skills? What about MCP? Subagents? Slash Commands? Let's break down EXACTLY when to use agent skills vs MCP servers vs sub-agents vs custom slash commands. 🔥 Here's the truth: Claude Code has evolved from a simple tool into a BEAST with agent skills, sub-agents, custom slash commands, output styles, plugins, hooks, and MCP servers. It's getting confusing AF. After a WEEK of wrestling with agent skill issues, I'm breaking down the RIGHT way and the WRONG way to use these features so you don't waste time like I did. 🎥 VIDEO REFERENCES - Claude Code Skills: https://docs.claude.com/en/docs/claude-code/skills - Multi-Agent Observability Codebase (get the meta-skill): https://github.com/disler/claude-code-hooks-multi-agent-observability - Tactical Agentic Coding: https://agenticengineer.com/tactical-agentic-coding?y=kFpLzCVLA20 In this video, I'm walking you through the REAL differences between agent skills, MCP servers, sub-agents, and custom slash commands. We'll dive deep into agentic coding patterns, showing you exactly when to use each Claude Code feature and why it matters for your AI coding workflow. I'll show you the wrong way engineers are approaching agent skills (converting ALL slash commands to skills - BIG MISTAKE), and then reveal the right compositional approach that leverages prompt engineering as your foundation. Watch as I build a Git worktree manager skill live, demonstrating how to properly compose custom slash commands, sub-agents, and MCP servers into a powerful reusable agent skill. This is tactical agentic coding at its finest. Whether you're deep into agentic engineering or just getting started with Claude Code and indydevdan's approach to AI agents, this breakdown will save you HOURS of confusion and help you master the core four: context, model, prompt, and tools. The prompt is STILL the fundamental unit of knowledge work. Don't give it away to complex abstractions. Skills are powerful, but they're compositional units - NOT replacements for your custom slash commands and prompt engineering fundamentals. We'll cover: - When agent skills shine vs when they're overkill - Why prompt engineering remains your most critical skill - How to compose features without circular confusion - Real production patterns for agentic coding workflows - The pros and cons of Claude Code's agent skills feature This is essential viewing for any AI coding engineer who wants to stay ahead in agentic engineering. Stop the confusion. Master the features. Ship better code. Stay focused and keep building. #aicoding #claudecode #aiagents

The first really great video IMHO about Claude Codes Skills vs. Slash Commands vs Subagents:

www.youtube.com/watch?v=kFpLz...

Kudos to @IndyDevDan!

huggingface.co/spaces/Hugging... #llm

Kind of cool: add #codex as an #mcp server in #Claude and use it in prompts as an independent reviewer:

claude mcp add codex -s user -- codex -m gpt-5 -c model_reasoning_effort="high" mcp-server

I use that all the time to review code by a different model than the writer.

Code like a surgeon

Great AI coding comparison: "A surgeon isn't a manager, they do the actual work! But their skills and time are highly leveraged with a support team that handles prep, secondary tasks, admin. The surgeon focuses on the important stuff they are uniquely good at."

www.geoffreylitt.com/2025/10/...

Great overview of Ruby Async/Threadsafe Antipatterns:

socketry.github.io/async/guid...

BTW, Claude Code + Codex are really good finding and eliminating those patterns in your codebase. Give it a spin…

Need a quick refresher? pawelurbanek.com/rails-thread...

Define tools in ruby_llm to be used for both tool calling as well as MCP

github.com/afstanton/tool_forge

Detect Electron apps on mac where the Electron hasn't yet been updated to fix the system wide lag Detect Electron apps on mac where the Electron hasn't yet been updated to fix the system wide lag - Detect electron apps causing macOS Tahoe lag.md

If you are running MacOS Tahoe and your system is lagging, it might be because of outdated electron apps.

tl;dr fix: Enter `launchctl setenv CHROME_HEADLESS 1` after every reboot.

More Info including snippets who find out which apps are the culprits:

gist.github.com/tkafka/e3eb63...

GitHub - ChromeDevTools/chrome-devtools-mcp: Chrome DevTools for coding agents Chrome DevTools for coding agents. Contribute to ChromeDevTools/chrome-devtools-mcp development by creating an account on GitHub.

Chrome #MCP DevTools are awesome to debug complex JS stuff in a browser: github.com/ChromeDevTools/chr... (announcement: developer.chrome.com/blog/chr...). Installation takes 2 minutes and works out of the box.

#ai #coding

Great write-ups about how #turbo actually works:

ducktypelabs.com/how-does-tur...

ducktypelabs.com/how-does-tur...

#rails #hotwire

Claude Code Docs, Guides & Best Practices | ClaudeLog Claude Code documentation, tutorials & best practices from real-world usage. Expert insights, optimization techniques, and searchable knowledge base.

Awesome resource about techniques how to use Claude Code: claudelog.com/

#AICoding #AI #Coding

#RailsWorld 2025, here we go!

NYC’s Urban Textscape Analyzing All of the Words Found on NYC Streets

Now that's cool applied machine learning and data science: pudding.cool/2025/07/stre...

🚀 Hiring now: Senior Ruby on Rails Developer | AI/LLM · DynamoDB · OpenSearch | Remote-Full time Employee No. | Thomas Witt 🚀 Hiring now: Senior Ruby on Rails Developer | AI/LLM · DynamoDB · OpenSearch | Remote-Full time Employee No. 2 | UTC - UTC +4 We’re veteran founders and angel investors who’ve seen too many terrible...

🚀 Hiring: Senior Rails Dev (AI · DynamoDB · OpenSearch) Employee No. 2

- Rails 8, Hotwire, Tailwind
- Live product, real customers
- Tiny senior team, daily shipping
- Remote (UTC–UTC+4), equity options

DM + GitHub👇
www.linkedin.com/feed/update/...

#rails #jobs #ruby #remote

My AI Skeptic Friends Are All Nuts My smartest friends have bananas arguments about LLM coding.

fly.io/blog/youre-a...

Word.