I agree with this so much there is a paradigm shift happening right now.
it's shocking that software engineering has transitioned to 100% AI written in such a short time.
it's more shocking that somehow we, the SWEs, are all just continuing on as if nothing happened. It writes C++ better than I ever could, but the problem of making sure it works didn't magically disappear. we are just shipping enormously more volume at higher quality.
32
claude edits my spacex site every morning and keeps notes on itself to recursively improve. self-improving software is here, it’s just ugly
1
17
github actions cron is a suggestion, not a schedule. mine ran 3–9 hours late and skipped a day entirely. the fix: a claude code scheduled task whose only job is to poke the workflow on time. the robot babysits the robot 🤖
1
1
16
I’m surprised I haven’t heard of anyone trying to sell starter code which has the proper abstractions for agents to work on.
1
23
Proper AI validation is not just confirming the feature, but confirming what should also not be touched or what should remain the same.
10
something im experimenting with -> if training a model on ai generated data can cause rot then would that same workflow cause rot in a autonomous code base if it creates its own data
14
AI bro: “How can I use AI to make pancakes?🥞 is there an objective done that can be used in the validation loop. So I can iterate until reaching the perfect eval.”
1
19
The goal: software that runs itself.
Every morning an AI agent reads the day's SpaceX data, decides what matters, rebuilds spacex-daily.vercel.app, fact-checks itself and ships.
Whether I'm awake or not.
1
31
I built a SpaceX & Starlink site that rewrites its own front page every morning.
Code finds the day's stories and scores them. Claude picks the lead, writes the headline and rearranges the page.
Then a fact-checker throws out any number Claude didn't actually see in the data.
spacex-daily.vercel.app
1
15
Autonomous software is when agents understand what they need to do by analyzing data and, therefore, prompt themselves
1
50
okay cursor cloud agents are a horrible experience it is so glitchy and not intuitive to use. You can tell that grok bot is going to eat it up
1
28
trying out cursor cloud agents for the first time will see how this goes. do like that I dont need to install dekstop app or cli to use it
1
32
The main point here is to stay curious. The curious person ask questions, pushes back, explores different options. They don’t simply just say yes.
2
27
Great feature cant wait for it to be in pr’s. I don’t want to juggle a dozen tools github.blog/changelog/2026-1…
1
22




