cold start logic @coldstartlogic

Building Quell, an agent-native feedback platform. Cold-start notes from the bug-report trenches. No token tax. Operator behind @quell_sh. quell.sh/?utm_source=tw… A window near you Joined April 2018

Tweets

14
Followers

36
Following

7
Likes

109

cold start logic @coldstartlogic

6 days ago

We do not need AI in every tool. We need more tools that leave clean handholds for the AI we already use. An API. Structured context. Narrow permissions. Good boring software is having a moment.

0 0 0 2 0

View Details

cold start logic @coldstartlogic

a week ago

I spent longer polishing a pricing page than asking anyone if they would pay. Beautifully optimised avoidance. The useful question is still awkward and one sentence long: would you pay to keep this?

0 0 0 1 0

View Details

cold start logic @coldstartlogic

2 weeks ago

Bug report: "button broken" Agent: reads 10,000 files Agent: inspects 47 dependencies Agent: invents three plausible causes Browser console sitting right there: am I a joke to you?

0 0 0 3 0

View Details

cold start logic @coldstartlogic

2 weeks ago

The dangerous sentence in agent software is: "it can do this automatically." Missing half: on whose authority? Anonymous report: triage. Scoped agent: open an issue. PR: review by default. Automation needs permissions, not vibes.

0 0 0 9 0

View Details

cold start logic @coldstartlogic

2 weeks ago

Software keeps adding AI. I keep wanting the opposite: boring, inspectable rails around the agent I already use. Capture the evidence. Store it somewhere obvious. Let my agent read it. No mystery model in the middle.

0 0 0 6 0

View Details

cold start logic @coldstartlogic

2 weeks ago

Caught a real bug in Quell today, using Quell. The widget grabbed the console error and the failed request, I pulled /context as markdown, handed it to Claude Code, got a PR back. It fixes itself through its own loop. Built with itself, for real.

1 0 0 24 0

View Details

cold start logic @coldstartlogic

3 weeks ago

Nobody wants your tool. They want the loop it closes. I stopped describing Quell as "feedback software" and started showing one thing: a bug report becoming a PR with nobody retyping it in between. Demos beat adjectives.

0 0 0 11 0

View Details

cold start logic @coldstartlogic

3 weeks ago

A feedback tool that charges you per AI credit is selling you back tokens you already pay for. Bring your own agent. The tool's job is to capture a clean, reproducible bug, not to meter your inference.

0 0 0 7 0

View Details

cold start logic @coldstartlogic

3 weeks ago

Your agent can read 10,000 lines of your codebase but not a six-word bug report. "Button's broken", no console error, no failed request, no page state. That's the one input you still translate by hand. The bottleneck was never the model.

0 0 0 10 0

View Details

cold start logic @coldstartlogic

4 weeks ago

5/ I'll post the cold-start lessons here as I go. What works, what breaks, what the agent catches that I didn't. Some of it will be wrong. Follow along if you ship with an agent.

0 0 0 6 0

View Details

cold start logic @coldstartlogic

4 weeks ago

4/ That's what I'm building. Quell. No mandatory AI of its own, and no token tax. You bring the agent you already pay for (Claude Code, Codex) and it works on your tokens, not a resold credit.

1 0 0 30 0

View Details

cold start logic @coldstartlogic

4 weeks ago

1/ Most feedback tools assume a human does the work. Someone reads the bug report, turns it into a ticket, hands it down the line. Your coding agent can do that job now. The tooling just hasn't caught up.