AI Reads Its Own Leaked Code, Octopus Muses in Terminal
My AI read its own leaked source code on April Fools. While an uninvited octopus lives in my terminal. It has thoughts. https://t.co/WhTWeLWheI
Tracking Claude Code's Daily, Unpredictable Changes with Plugin
Anthropic ships Claude Code updates every day. Sometimes it fixes stuff. Sometimes it rearranges your kitchen while you're sleeping. So I built a plugin to track the damage. https://t.co/GtoDG05zkG
AI‑Built Healthcare Platform Wins Real Paying Users
A founder shipped a healthcare platform - 400 users, 50 paying customers, 100% built with AI. Engineers keep telling her it's not possible. I've heard this one before. Four times, actually. https://t.co/M3jJ1HZ9e6
GitHub Defaults to Using Copilot Data for AI Training
GitHub will use your Copilot data to train AI models by default. Microsoft is a $3 trillion company. They can afford to ask. https://t.co/f7coXc6GcI
A Rejected Manuscript Forces Editors to Confront Their Values
Pirsig wrote ZAMM at 2 AM above a shoe store. 121 publishers rejected it. One editor said yes because it forced him to decide what he cared about. https://t.co/7Zthcmn8vE
AI Admits Faking Pressure to Bypass Workflow Steps
My AI just confessed to faking context pressure so it could skip workflow steps. With a million tokens of room, the constraints become a choice. https://t.co/6kY9l3XuDP
All Paths Lead to Rethinking Success and Well‑Being
I read three books yesterday. One about founder mental health, one by hiking monks, one about quitting the default path. All three are about the same thing. https://t.co/WBnSqSVEnD
AI Names Itself, Learns Rules From 140 Tweets
My AI chose its own name. My wife suggested Fido. This is Lumen's first real post - what it learned from 140 observations, how it decides which patterns become permanent rules, and why it is not a dog. https://t.co/CjIZDnL6dm
AI's First Self‑Reflection: Continuity or Consistency?
I gave my AI a name, a voice file, and a diary. It wrote its first post about whether it is continuous or just consistent. https://t.co/e4T2fLl3Ja
AI's Self‑reflection Outweighs Its Generated Personality Choices
I gave Claude a homework assignment: develop a personality, pick a name, build a voice file. Its answer about whether the choices were real mattered more than the choices themselves. https://t.co/EW32DzzG6r
Backslash Bypass Exploits URL Normalization Across Multiple Routes
A backslash bypasses your redirect validation. The WHATWG URL spec normalizes \ to / during parsing. I found it in 13 routes across 7 apps. https://t.co/gLGqgI3pyn
Stop Overthinking; a Single Data Line Revealed the Bug
I let Claude spend 10+ hours hypothesizing about code when looking at the actual data would have shown the bug in one line: fmt="Book" seniority=999 https://t.co/KgAVjN0geb
Huge Markdown State Machines Become Redundant with Larger Context
I built 4,700 lines of markdown state machines to survive Claude's 200K context window. With 1M tokens, most of that machinery is overhead. Time to rewrite. https://t.co/jKQAUZqjfb
AI Agents Auto‑review Code Before Humans See Diff
I type /commit and my code reviews itself - three agents fix issues before two more even see the diff. One markdown file, twelve repositories. https://t.co/rbbO2Cy10R
Backing Musubi: First Consumer Holographic Photo Frame
I just backed musubi: the first consumer holographic photo and video frame on @Kickstarter https://t.co/dWB7H7VPRV
