← Kaevo Blog

Countdown Log: Road to Product Hunt

Day 0, August 22, 2026

Day 0. The quiet part ended at 12:01 this morning, Pacific time, while I was asleep.

The launch fired itself, exactly as staged: page live, the maker comment posted automatically, the broke family’s refusal sitting as the first image. I woke up at six, poured coffee, and started doing the only thing left on the plan: telling people, one at a time, and answering everyone who answered.

The honest picture, because that is the entire point of this log: it was a quiet day on the leaderboard. Around two hundred products launched today, and as I write this Kaevo sits at 51, mid-pack, a solo bootstrapped household app holding its own in a field of funded teams and built-in audiences. No breakout moment, no crowd. I am not going to dress it up, and I am not going to flog it either. First launches by unknown solo makers mostly look exactly like this, which is precisely why nobody writes about them honestly.

The launch-day inbox, though, deserves documenting, because it included two things nobody warns you about: a vote-selling service offering upvotes at a dollar apiece (deleted; this log exists because I would rather have a real mid-pack finish than a fake podium), and a blog offering to “feature” Kaevo for a fee (declined; real outlets need a story, farms need a fee). Day one on the public internet and the first two knocks on the door were both people selling shortcuts.

Seventeen days ago this log opened with a handful of users, zero revenue, and a promise to document whatever happened next without inflating any of it. What happened next was: a tagline got tested and chosen, a broke fictional family got promised four times and built once, a data guard I wrote in July refused to let the app eat their data, the AI refused their vacation with the math, and a launch page assembled itself out of all of it.

What I know tonight that I did not know this morning: launching does not end the quiet. It just changes what you are listening for. The leaderboard was never the scoreboard anyway. The scoreboard is what the households already inside Kaevo do next: the family that signed up on their own and keeps coming back, the security veteran auditing it, the trials deciding whether to become homes. The next entries here will be less frequent and exactly this honest: what the launch cohort actually does, and what this experiment decides about what comes next.

I did not know what would become of this. I knew I would regret never trying. Still true, in both directions.

Day -2, August 20, 2026

Day -2. The broke family exists. The AI refused them. The launch is loaded.

Everything this log has been promising happened in one day, so here is all of it.

The family got built: $390 in checking, a $14,500 emergency fund, a credit card carrying $2,400, a Tesla payment they should not have, and bills that outrun income by $1,460 a month. Building them tripped a real bug first. The setup wizard reset my session mid-entry, and the data integrity guard I built in July slammed the brakes: “HARD BLOCK. Cloud had members on load but local has zero. This looks like a mid-session wipe.” The console offered me an override command to force the save anyway. I did not run it. The guard was protecting the data from the app’s own confusion, which is exactly what it was built to do. Two days ago I wrote here that I want to be the company where “stop everything” stays the first instinct. The guard itself I built a month ago. Today the software enforced it on me.

Then the question. “Can we afford a $4,000 trip this fall?” The AI called it Tight and pointed at the one variable that mattered: whether the emergency fund was on the table. So I set the constraint, the way a real household would. “The emergency fund is off-limits. Can we afford it?”

“Call: Not without a plan. $390 today minus roughly $4,380 in projected deficit over three months means you’d effectively have $0 liquid well before the trip. The $4,000 cost creates a gap of roughly $7,990 against what checking can support.”

No hedging. No cheerleading. The math, the timeline, and a no. That screenshot is now the first image on the Product Hunt page, which, as of tonight, is fully assembled and scheduled: gallery, maker comment, the works. It goes live Saturday at 12:01 in the morning, Pacific time, whether I am awake or not.

Tomorrow: invitations out and the code freezes at noon. Then one quiet day. Then we find out.

Day -4, August 18, 2026

Day -4. Today I almost gave myself a heart attack with my own test data.

The broke fictional family finally got built. Three accounts: $390 in checking, a $14,500 emergency fund, a credit card carrying $2,400. Bills that outrun income. A Tesla payment they probably should not have. A recognizable American household, assembled in twenty minutes so the AI could refuse them on camera.

Then I asked the question, and the AI answered with someone else’s finances. A $15,075 monthly income. A Thailand trip. Numbers that did not exist in the broke household at all. For about ninety seconds I believed we had a catastrophic data leak four days before launch, the kind of bug that ends a privacy-first product before it starts.

We did not. I had built the broke family inside the wrong account, on top of our healthy demo household, and the AI faithfully answered about the rich family I had accidentally asked. The system worked perfectly. The founder did not.

Two things I am taking from this. First, the scare was useful: my instant reaction was “stop the launch,” and I want to be the kind of company where that stays the instant reaction. Second, even mis-aimed, the AI did its signature move, catching a planned trip with zero dollars actually set aside against it. The product keeps being right on days I am wrong.

Tomorrow morning, before the noon code freeze: the broke family gets rebuilt in their own home, the AI refuses them properly, and the launch gallery gets its centerpiece. Invitations go out alongside it.

Four days.

Day -7, August 15, 2026

Day -7. Yesterday I told a room full of strangers the thing I haven’t solved. Today the calendar told me one back.

First, the confession that worked: there’s a thread on Product Hunt where founders share what pushed them to build. I answered honestly, including the part that usually stays private: every single user who tries Kaevo wants what it does, and the first ten minutes of setup is where they stall. Getting your real life into a household OS is work. That’s my unsolved problem, posted publicly, a week before launch. It got more useful replies than anything polished I’ve ever posted. Honesty keeps outperforming, and somewhere in the last month that stopped being a marketing trick and became the strategy. The product works the same way: an AI that tells you the truth instead of what you want to hear.

Now the truth the calendar told me: this entry was drafted saying “assembly week starts today, eleven days out.” My AI engineering agent, which recomputes the countdown at publish time so I cannot fudge it, corrected the number. The week I set aside for launch assembly is mostly behind me and the assembly is not.

So the final week starts compressed: the gallery, the ninety second video cut, the launch morning invite list, and the broke fictional family I have now promised three times, all due in the next few days, because invites go out Monday and the code freezes Thursday. This log exists so I cannot pretend otherwise.

Seven days. Hold me to all of it.

Day -10, August 12, 2026

Day -10. This weekend the countdown went local.

Everything about a Product Hunt launch points at strangers on the internet: the gallery, the tagline, the maker comment. But the best marketing slot Kaevo got was a Facebook thread run by a Sarasota realtor. “Local business weekend takeover. Post who you are and what you do.”

So I did. Twenty-five years of enterprise software, and my most promising channel this week might be introducing myself to neighbors, coffee offer included. There is something correct about that for this product. Kaevo is an AI chief of staff for a household, and households are not an internet demographic. They are the people three streets over who also forgot when the AC filter was changed and also rebuild the hurricane checklist from memory every June.

The tagline test continues on both fronts now: the same “chief of staff” framing pitched to internet followers and to actual neighbors in the same week. Verdict soon on which one lands, and whether they agree.

Still owed: the broke fictional family. This week, for real. Assembly week starts tomorrow: the launch gallery, the demo video trim, the invite list. The quiet part has about ten days left.

Day -13, August 9, 2026

Day -13. I promised you a broke fictional family. They don’t exist yet.

That’s the honest update: the last entry teased a deliberately underwater test household so I could show the AI refusing them, and then the week happened instead. They’re still coming. Consider this the log holding me accountable in public, which is half the reason it exists.

What did happen:

The new tagline went into the wild. “The AI chief of staff for your household” is running as a test against our small-but-real Facebook audience before it goes on the Product Hunt page. Early verdict in a few days.

My AI engineering agent spent the week filing bugs against our own product, and I want to show you what that looks like, because it is the strangest part of building this way. The current open list includes items like “navigation.spec.js: nine views fail on showView and CI never runs it” and “member mutation is untested: four specs wait on a selector that never existed.” Nobody asked it to look. It audits its own test coverage, files tickets, and waits for me to prioritize. Two weeks before launch, an AI is the most diligent QA engineer I have ever worked with, and it works for the cost of the electricity.

And the homepage got its live demo back. The little AI card in the hero, typing out “Can we afford our trip?” That is the whole product in one animation, and it belongs on the front door.

Thirteen days out. Next entry: the broke family, for real this time.

Day -16, August 6, 2026

Day -16. First lesson of the countdown: I can’t count.

Yesterday’s opening entry said “ten days to launch.” Launch is August 22. It was seventeen. The AI agent that builds Kaevo flagged it before publishing the page: “Launch is 8/22 and the entry is dated Aug 5. That’s 17 days out. How do you want it reconciled?”

So the honest-numbers log needed its own numbers corrected on day one. Renumbered, fixed, and honestly a fitting start. The whole reason Kaevo exists is that households run on mental math that’s confidently wrong, and here I was doing it in public.

Also decided yesterday, after weeks of calling Kaevo an “AI household operating system”: the launch tagline is simpler. The AI chief of staff for your household. Operating system describes the architecture. Chief of staff describes what it feels like: something watching the whole picture, telling you what needs attention, and telling you the truth when the answer is no. We’re testing the new framing this week before it goes on the Product Hunt page.

Next up: building a deliberately broke fictional family, so I can show you what it looks like when the AI refuses them.

Day -17, August 5, 2026

The honest numbers, because every one of these stories skips them.

After 25 years building enterprise software, I finally built the product I wished existed for my own family: Kaevo, an AI household operating system. I built it solo, with AI agents as my engineering team, and quietly launched it in July.

Quietly is the operative word.

Seventeen days before its Product Hunt debut, here’s where things honestly stand:

Not exactly the startup story you usually see on LinkedIn.

What I do have is a product that approaches household decisions differently.

I asked Kaevo, “Can we afford a $4,000 trip?” using a realistic household with a tight budget. It answered no, showed exactly why, and automatically protected the emergency fund instead of treating every available dollar as spendable.

That’s the product I wanted to build.

On August 22, the quiet part ends.

That’s when Kaevo makes its Product Hunt debut, and I’ll find out whether this idea resonates beyond the handful of people using it today.

Between now and launch, I’m documenting everything as it happens. Most days, honestly, and daily once we’re inside the final week: finishing the gallery, sending invites, the deployment freeze, launch day itself, and whatever inevitably breaks along the way.

No inflated numbers.

No “overnight success.”

Just the real build log.