Co-Founder @reticle_hq
Making AI coding agents deliver || Saving humans from becoming a feedback loop for AI
IIT & IIM Grad || Living life with Curiosity
The most dangerous response in AI coding might be 200 OK.
Your agent sees success.
You see a working feature.
Then you refresh the page.
The data is gone.
200 OK proved the server answered.
Nothing more.
We shipped v2.9.0 today that makes our product do less.
If you use AI coding tools, every extension you add competes for the same limited slot count in your editor. We were taking up 48 slots. Installing us could quietly push someone's other tools out.
We cut it to 18. Nothing was removed, it just stopped shouting for attention.
You don't earn a spot in someone's workflow by taking up the most room in it.
If you use AI coding agents, try @reticle_hq now.
my bio says 'saving humans from becoming a feedback loop for AI.' I mean it literally.
an agent writes a change, reviews its own work by looking at its own output, declares it done, and loops. The human is just the break condition - we step in when something feels off.
That's not verification. That's a human-in-the-loop alarm clock.
The fix isn't a smarter agent. It's ground truth the agent can read but can't influence - the running app's network, store, console - checked after the action settles, before a human ever opens the tab.
That's why we're building Reticle. The longer bet: verification belongs in the agent loop itself, not bolted on after, and the team that makes it a primitive, not a feature, sets the category.
@sridharfyi this is interesting. kudos to you.
FYI, we have built a verification layer for coding agents. I believe this will be interesting for you. Lets chat sometime