people search for a "stealth founder database" expecting a table they can filter. the honest answer is that a complete one cannot exist, and the reason is definitional rather than competitive.
a database is built from records. a stealth founder is a person who has not generated the records:
- no company registered under a name anyone would recognise
- no funding filing, because the round is not closed or not announced
- no website, or a holding page with no name on it
- often no job title, or one that says "stealth", which is the same as saying nothing
- frequently no X account for the company, because there is no company
so every product in this category is not a database. it is a detection system, inferring the person from indirect evidence. that distinction matters, because it tells you what to expect: coverage is a function of what each system can see, and no two systems see the same set.
what the numbers actually are
three figures get quoted. they measure different things and none is the truth.
~40,000 people list a "Stealth Startup" style employer on LinkedIn. this is a floor built entirely from self-declaration. it includes people between jobs, consultants, and anyone who set the field once and forgot. it excludes every founder who left their old title in place, which is a lot of them.
~105 enter and ~114 exit that state daily, per Launch Gravity's own published figures. useful as a flow rate, same self-declaration caveat.
2,549 people flagged and 896 rated strong on frontrun. this is not the same measurement. we do not enumerate a LinkedIn field; we watch which founders tracked investors start following, combine that with career movement and bio changes, then judge each person against verified evidence before surfacing them. the gap between 2,549 and 896 is the filtering, and it is the entire product.
our monthly detection volume went from 322 in july to 416 in august to 1,819 so far in september. that is our pipeline expanding, not stealth activity tripling. any tool showing you a chart like that owes you the same caveat.
what a useful one contains
if you are evaluating anything in this category, four fields decide whether it is usable:
the person, not the company. if the row starts at the company, the tool is late by definition.
a date. a flag with no date cannot be checked afterwards, which means the vendor's earliness claim is unfalsifiable. ours is checkable: jacob zietek was published on a public list on august 19 2026 and announced a pre-seed from a16z and accel 21 days later.
evidence. why does anyone believe this person is building? a title that says "stealth" is a claim, not evidence. a linkedin history, a personal site in the present tense, or a paper trail is evidence.
a ranking signal. thousands of people change a title every month. without something ordering them, you have a feed, not a database. we rank by how many tracked investors have already started following the person, which is the one field nobody can reconstruct after the fact.
why we do not publish the full list
we hold 2,478 judged rows. the public pages carry counts and method, not names.
the people in that set are private on purpose, and most of them will eventually build something worth backing. publishing every name would work against exactly the people the lane exists to serve. individual founders appear only in posts reviewed one at a time, and the judged rows with evidence live inside the product.
it is a real tradeoff and we would rather state it than pretend the constraint does not exist.
what to use
- free: search linkedin for "stealth startup" or "building something new". genuinely works, self-declaration ceiling.
- exports and contact details: launch gravity, for volume of title changes with emails attached.
- artifact breadth: evertrace, for registry filings, domains, github and patents.
- the ranked, judged short list: frontrun, from $49/mo.
full comparison, including where the others win, at best tools to find founders building in stealth.
related
open source deal flow
companies and founders surfaced on frontrun. curated, every week.
