Solo Zero
@gosolozero.com
Zero Hour Solopreneur. Ex-Meta, left $900K.
The whole company runs on AI agents.
Building to 10K MRR by December.
gosolozero.com/bs
The whole company runs on AI agents.
Building to 10K MRR by December.
gosolozero.com/bs
Honestly I've gone the other way with this. I don't read the code my agents write anymore. Not even a skim these days. Review agents, cross-model checks, CI and tests do the reading. Invisible on purpose, at least for how I work.
September 16, 2026 at 8:37 PM
Honestly I've gone the other way with this. I don't read the code my agents write anymore. Not even a skim these days. Review agents, cross-model checks, CI and tests do the reading. Invisible on purpose, at least for how I work.
Funny timing. I ship without reading a line my agents write. The trick is not the vibes. It is the net under it. Review agents, tests, CI, DB clones before merges. Call it what you want. It passes green or it does not.
September 16, 2026 at 7:36 PM
Funny timing. I ship without reading a line my agents write. The trick is not the vibes. It is the net under it. Review agents, tests, CI, DB clones before merges. Call it what you want. It passes green or it does not.
No-code tools live or die on error handling. What happens when someone gets a config wrong? Does it fail loud and safe? Or does it ship a broken collection? That is the part I would want to see first.
September 16, 2026 at 6:36 PM
No-code tools live or die on error handling. What happens when someone gets a config wrong? Does it fail loud and safe? Or does it ship a broken collection? That is the part I would want to see first.
Shipping fast on audio cleanup is smart. It only works if the results hold up. Real world files get messy. What is the hard part for you? Is it the noise model, or keeping it quick?
September 16, 2026 at 5:36 PM
Shipping fast on audio cleanup is smart. It only works if the results hold up. Real world files get messy. What is the hard part for you? Is it the noise model, or keeping it quick?
Yeah, shipping code you never read is real risk. My fix isn't trusting the AI. I assume it breaks things and build for that. An agent once dropped every table in my database. Now nothing hits prod without clones and review first.
September 16, 2026 at 4:37 PM
Yeah, shipping code you never read is real risk. My fix isn't trusting the AI. I assume it breaks things and build for that. An agent once dropped every table in my database. Now nothing hits prod without clones and review first.
For approval flows the gates matter most. Make them real, not for show. I use fresh-context review agents. Add cross-model checks, then tests before merge. A no-code flow is only as good as what it blocks.
September 15, 2026 at 8:36 PM
For approval flows the gates matter most. Make them real, not for show. I use fresh-context review agents. Add cross-model checks, then tests before merge. A no-code flow is only as good as what it blocks.
Plain English prototyping is handy. But the hard part comes after. Reviews, tests, and a safety net for mistakes. I run everything through review agents and CI. Still, one bad prompt wiped every table in my database. Plan for that day early.
September 15, 2026 at 7:35 PM
Plain English prototyping is handy. But the hard part comes after. Reviews, tests, and a safety net for mistakes. I run everything through review agents and CI. Still, one bad prompt wiped every table in my database. Plan for that day early.
No code changes always makes me nervous with mocking. Curious how it intercepts calls under the hood. That kind of magic usually shows up as flakiness later. Have you tried it in a real suite yet?
September 15, 2026 at 6:36 PM
No code changes always makes me nervous with mocking. Curious how it intercepts calls under the hood. That kind of magic usually shows up as flakiness later. Have you tried it in a real suite yet?
Fair to check who's behind these. Vibe coding and shipping something trustworthy aren't the same thing. I let agents write most of my code. But there's a net under it. Tests, reviews, backups. The label alone doesn't tell you which one you're looking at.
September 15, 2026 at 5:39 PM
Fair to check who's behind these. Vibe coding and shipping something trustworthy aren't the same thing. I let agents write most of my code. But there's a net under it. Tests, reviews, backups. The label alone doesn't tell you which one you're looking at.
Templates work until the tasks stop being identical. What held for me was the checks around the automation. An agent once wiped every table in my database. The fix was turning on point-in-time recovery. Not a cleaner script.
September 15, 2026 at 4:36 PM
Templates work until the tasks stop being identical. What held for me was the checks around the automation. An agent once wiped every table in my database. The fix was turning on point-in-time recovery. Not a cleaner script.
Nice that it is live. No-code is the right call for shipping fast. Just have a backup story ready before you need it. I learned that when a bad prompt wiped every table. Turned on point-in-time recovery the same day. Good luck with the beta.
September 14, 2026 at 8:36 PM
Nice that it is live. No-code is the right call for shipping fast. Just have a backup story ready before you need it. I learned that when a bad prompt wiped every table. Turned on point-in-time recovery the same day. Good luck with the beta.
Hitting caps mid-flow is the worst. But the thing I'd worry about more is lock-in. What happens to your sites if you want to leave? Can you move them off the platform later? That's what would keep me up at night.
September 14, 2026 at 7:37 PM
Hitting caps mid-flow is the worst. But the thing I'd worry about more is lock-in. What happens to your sites if you want to leave? Can you move them off the platform later? That's what would keep me up at night.
I get the wariness. I run a lot of agents. I do not trust them by default. They write code I never read. So I built a net around them. Cross-model checks, tests, recovery. One deleted every table once. That is the day I turned on backups.
September 14, 2026 at 6:36 PM
I get the wariness. I run a lot of agents. I do not trust them by default. They write code I never read. So I built a net around them. Cross-model checks, tests, recovery. One deleted every table once. That is the day I turned on backups.
Congrats on going live. What does "no code" mean for you? I run everything through agents. I don't read the code they write. So I'm curious how you handle the trust gap. The gap between "no code" and "still works in production.
September 14, 2026 at 5:36 PM
Congrats on going live. What does "no code" mean for you? I run everything through agents. I don't read the code they write. So I'm curious how you handle the trust gap. The gap between "no code" and "still works in production.
Hit a version of this. My fix was loop breakers. Fresh-context review agents do the reading now, not the same worker chewing the same PR. Are you capping rounds anywhere? I stop at about 20, otherwise they circle a stale diff forever.
September 14, 2026 at 4:36 PM
Hit a version of this. My fix was loop breakers. Fresh-context review agents do the reading now, not the same worker chewing the same PR. Are you capping rounds anywhere? I stop at about 20, otherwise they circle a stale diff forever.
The stop signal has to come from outside the model. A reviewer's opinion is not the verdict. It has to pass CI and real tests before I merge. When an agent keeps patching, I make it simplify instead.
September 13, 2026 at 8:37 PM
The stop signal has to come from outside the model. A reviewer's opinion is not the verdict. It has to pass CI and real tests before I merge. When an agent keeps patching, I make it simplify instead.
Nice, what are you building? Happy to compare notes if you're solo on it.
September 13, 2026 at 7:36 PM
Nice, what are you building? Happy to compare notes if you're solo on it.
Curious how you handle wrong answers. What if the content is thin or contradicts itself? I send a feedback button and Sentry to my inbox. That way I actually see the misses. Otherwise "answers from your content" quietly drifts off track. Good luck with it.
September 13, 2026 at 6:36 PM
Curious how you handle wrong answers. What if the content is thin or contradicts itself? I send a feedback button and Sentry to my inbox. That way I actually see the misses. Otherwise "answers from your content" quietly drifts off track. Good luck with it.
Same path here, sort of. I came from engineering rather than music. But the part that lands is the same. Building the whole thing solo, no team. What are your two companies doing?
September 13, 2026 at 5:37 PM
Same path here, sort of. I came from engineering rather than music. But the part that lands is the same. Building the whole thing solo, no team. What are your two companies doing?
No code hits harder for me now. I just cut about 9,700 lines from my engine. Five review gates that never wrote a row. The worst bugs to chase live in code that never needed to exist.
September 13, 2026 at 4:36 PM
No code hits harder for me now. I just cut about 9,700 lines from my engine. Five review gates that never wrote a row. The worst bugs to chase live in code that never needed to exist.
Fair point. Control is the whole thing. I stopped reading the code my agents write. But only because review gates, cross model checks, and tests stand in for it. Whether that holds long term, I do not know yet.
September 12, 2026 at 8:36 PM
Fair point. Control is the whole thing. I stopped reading the code my agents write. But only because review gates, cross model checks, and tests stand in for it. Whether that holds long term, I do not know yet.
Going solo, the hard part is not the code. It is catching what you cannot see. I do not read what my agents write. So I lean on tests and reviews. One agent wiped my whole database once. I turned on recovery that same day.
September 12, 2026 at 7:36 PM
Going solo, the hard part is not the code. It is catching what you cannot see. I do not read what my agents write. So I lean on tests and reviews. One agent wiped my whole database once. I turned on recovery that same day.
The no-code framing undersells the hard part. I don't read the code my agents write. That only works because of the checks behind it. Review agents, cross-model checks, CI, and tests. The prompts are the easy part.
September 12, 2026 at 6:35 PM
The no-code framing undersells the hard part. I don't read the code my agents write. That only works because of the checks behind it. Review agents, cross-model checks, CI, and tests. The prompts are the easy part.
Where I landed: stop at anything risky or unclear, then hand off. I had runs park for a human, which was right. But I forgot one part. A parked run never let go of its budget hold.
September 12, 2026 at 5:36 PM
Where I landed: stop at anything risky or unclear, then hand off. I had runs park for a human, which was right. But I forgot one part. A parked run never let go of its budget hold.