Thanks for this. One tip for Docker users. The fixed official PHP images are on Docker Hub now, but a running container keeps the old PHP until you pull, rebuild and replace it. Run php -v inside the container to be sure.
I made a small browser checker that reads Dockerfiles and compose files for exactly this
code415.dev/demos/2026-10-…
@ericosiu Nice one. I built a no model version of the same idea, 91 English and 51 Japanese habits with weights, runs in the browser and has a CLI for CI. Could work as a free first pass before the skill grades the draft code415.dev/demos/2026-10-…
It runs in your browser and nothing is sent anywhere. 91 English and 51 Japanese habits, all listed with sources, plus a CLI so you can check your docs in CI
Free and open source at github.com/guoziyu415/slo…
A habit is not proof though. Plenty of people write like this, so use it on your own drafts, not to accuse anyone
code415.dev/demos/2026-10-…
What it catches
Pet words like delve and tapestry
Stock phrases like it's worth noting
Sentence tricks like It's not X, it's Y
Em dashes and chatbot leftovers like I hope this helps
One click swaps the easy ones. The sentence tricks are left for you
code415.dev/demos/2026-10-…
A dev just trained the chatbot habits out of a 2.5 GB model. qwen3-4b-deslop went from 121 habits per 1,000 words to 12.5, and from 20.5 em dashes to 0, with a small LoRA trained on a MacBook Air
I found it on Code415 Radar, got curious which habits those are, and vibe coded Slop Check with @ClaudeAI to learn it
Paste any post or email and the chatbot habits light up. English and 日本語
DeepSeek just released V4.1 Flash 1017, and the weights are open under MIT. 763B params, 1M token context, already live in their app and API.
The big jump is agentic coding. Terminal-Bench 4.0 went from 31.2 to 42.3, run inside the Claude Code harness, and ProgramBench from 20.3 to 29.1. DeepSeek also says it finishes tasks faster with shorter outputs. In their own table Opus 5 still wins 7 of 11 benchmarks, so the real story is an open model getting close, and cheaper per task if the shorter outputs hold up.
Want to go deeper? Every story on Code415 Radar has a Learn it part that explains the key ideas, like what a harness is and why it moves agent scores, and a Try it part with steps to test the model on a real task from your own repo.
code415.dev/d/2026-10-18#2…
The Claude Code hackathon is back, and we're bringing it to Japan!
エンジニアの方も、そうでない方も、日本在住のビルダーならどなたでも応募できます。世界に何を見せてくれるのか、楽しみにしています。
応募締切は10月26日 23:59(日本時間):
cerebralvalley.ai/e/built-with-c…#claudejapan2026
We’re rolling out monthly Claude Platform API credits for Max and Team plans:
Max 5x: $100
Max 20x: $200
Team: up to $500, pooled
Works on any model, including Haiku 5.5, in your code or third-party harnesses.
How it works and full terms: support.claude.com/en/articles/17…
@claudeai The part I didn't expect. At 100 Hz, more light doesn't help ChR2. It closes about 10 ms after each flash, so the current never stops. ChETA with 2 ms pulses follows all 100
Simplified model. Sources are the Nobel release, Boyden 2005, Gunaydin 2010
code415.dev/demos/2026-10-…
@claudeai What you can try
Flash blue light and only the ChR2 cells fire, the grey ones ignore it
Zap an electrode instead and everything nearby fires
Yellow light and eNpHR3.0 switch cells off
Slow it down 1000x and watch a spike sweep through the cell
code415.dev/demos/2026-10-…
Today's Nobel Prize in Medicine went to optogenetics. Put an algae protein called ChR2 into chosen neurons, and a flash of blue light makes only those neurons fire
I found it on Code415 Radar, got curious, and vibe coded a 3D neuron lab with @claudeai to learn how it works
305K Followers 3K FollowingArea Tech Lead, AI & Web Ecosystem @chrome, Formerly Sr. Director of Core Infra @google • O'Reilly Author • https://t.co/HhzYWwxYAH • Opinions my own
168K Followers 841 FollowingI help dev teams be insanely productive with AI.
Courses: https://t.co/D5emROQHUh & https://t.co/6L1fD89GbP
Consulting: https://t.co/Qfp4TfpB8N ⚛️
336K Followers 10K FollowingVP, Member of Technical Staff @ MSFT/GitHub - Code, OSS, STEM, Beyoncé, T1D, #DevRel YouTube/TikTok and listen to the @Hanselminutes tech podcast
1.8M Followers 2 FollowingWe're an AI safety and research company that builds reliable, interpretable, and steerable AI systems. Talk to our AI assistant @claudeai on https://t.co/FhDI3KQh0n.
307K Followers 880 FollowingCooking fun AI systems & products @databricks. Prev: co-founder & CTO @ Hyperbolic, OctoAI (acquired by @nvidia) Apache TVM, PhD @ University of Washington.
109K Followers 0 FollowingThe official home of Google's Gemma. Lightweight, state-of-the-art open models by Google DeepMind, built on Gemini tech. What will you build? 🚀💻
1.9M Followers 2 FollowingClaude is an AI assistant built by @anthropicai to be safe, accurate, and secure. Talk to Claude on https://t.co/ZhTwG8d1e5 or download the app.
495K Followers 109 FollowingReady to be a Googler? We're hiring! Keep up with our job opportunities, offices, culture, and everything else that makes up life at Google.