i built a Claude Code style Markdown editor that runs in the browser https://t.co/oRaz4fU6Cs

if you find bugs, open issues in the flashtype repo 🫶 https://t.co/FwDH6SgOO6

flashtype is using https://t.co/AZpAXkLGuV to display the diffs, checkpoints, etc. lix is change control SDK i am building (think git but for apps & agents)

can't beat people having fun building things

must read for devs writing >90% of code with ai blast radius, detecting duplicate code, and ... just prompt. don't go crazy with mcp, sub-agents and what not 💯

getting closer to an alpha release. it's fast now. next up is a claude code like agent creating changes with the lovely accept/reject terminal ui 🤖

90% storage size reduction :OO had to optimize lix a bit more to get faster diffs in the markdown app this week. now continuing with the markdown app

I am building a claude code markdown editor in the browser goal for this week: get ready to launch it next week!

this is how i use TDD (test-driven development) with claude code: 1. chat about a new api 2. let claude code generate a test file with EMPTY tests 3. review the tiles of the tests 4. based on the titles of the tests, let claude code run loose this method ensures that claude

openai's codex cli with gpt 5 became better than claude code 🤯 it crawls the codebase to a degree i have never seen seen from claude code. Instantly one-shotted a bug i couldn't solve with claude code for 3 days new $200 per month subscription **check**

@meltyinc no brainer use case for dev tool companies: users that ask questions on discord. just copy & paste the thread with "clarify the docs" and et voila this comes out of it https://t.co/hvA6nmTW05

Conductor from @meltyinc is a game changer. background agents running locally without the annoying image bootup of cursor agents or open ai's codex. leads to way faster iterations

I offered everyone to come to the US, but nobody was able to commit - family, kids, personal stuff, all reasonable arguments Thus, we mutually separated last month :( What can be learned from this situation? 1. Distributed teams are disadvantaged in the age of AI. Async is

That's why I said "I was able to fire my team because of Claude Code". Claude Code + o3 were my sparring partners. I didn't need a team for architecture stuff. Was this ideal? Hell no. I had great talent in Europe, but I couldn't leverage it anymore

We decided to rebuild lix, which put me into an interim eng lead position AI meant that I was able to iterate faster with claude code + o3 on architecture stuff than relying on my team, which was 6-9 hours away. AI answers instantly without a multi-hour delay.

But we progressed too slowly. Git compatibility was identified as the root cause. We had two options: 1. Break up with git and rebuild lix from scratch, killing momentum for inlang. 2. Keep the focus on inlang at the risk of missing the opportunity for lix.

Everything went perfectly until mid-2024. We suddenly faced an innovator's dilemma situation. We built https://t.co/BGTSrgF1qc as a proof point for @lixCCS, a universal version control system: - any file format (not just text!) - SDK for apps - running in the browser

So what went wrong? In late 2023, I had to move to NYC to leverage US investors and customers. The team stayed in the EU. Which made sense. Most of our users for the GTM product were in the EU. Plus, we reached a point where the team operated independently of me.

Was it bad hiring? No. This team executed. Our GTM product ( https://t.co/BGTSrgF1qc) was loved by users, adopted among enterprises, and had a clear path to success.

I started outshipping my remote team with the emergence of AI. ChatGPT o3 for architecture. Cursor/Claude code to execute. I was shipping faster than ever.