# The Stealth Founder Resolution Teardown: Confirmed, Probable, or Not Forming *Start from one placeholder profile and finish with a written grade of Confirmed, Probable, or Not Forming that cites two independent records, names co-founders, and sets a dated decision.* - Canonical URL: https://www.refolk.ai/guides/stealth-founder-resolution-teardown - Pillar: Investing and deal sourcing - Format: Teardown - Published: 2026-10-22 - Last reviewed: 2026-10-22 - Reading time: 22 min - Keywords: how to find out what a stealth startup is building, identify stealth startup founders, stealth startup company name lookup, find co-founders of a stealth company, verify stealth founder before outreach ## Key takeaways - In Refolk's index of professional profiles, only 233 US profiles carry a Stealth Startup, Stealth Mode, or Building something new title, while about 38,000 people list the Stealth Startup page as their employer, so search the employer field, not the title. - Adding an Owner/Partner or CXO seniority filter to placeholder-title search in Refolk's index returned 0 profiles, so infer seniority from the prior role, not the current label. - A Delaware record with no officers proves nothing, because a corporation's first annual report is not due until the March 1 after formation and LLCs never file one; California's Statement of Information, due within 90 days of foreign qualification, is often the earliest public officer list. - More than 50% of about 40,000 venture rounds in one academic sample filed no Form D, so a missing Form D never means a stealth company is unfunded. - Run the co-departure scan before registry, domain, and filing lookups, because without candidate names and entity names the footprint searches have nothing to match against. - A resolution only counts when it cites at least two independent records and ends in a dated reach-now, watch, or drop decision; absences are logged but never counted as evidence. This is the procedure for one job. A high-signal operator's profile now says "Stealth Startup" or "Building something new", and you need to find out from public records whether a company exists, who else is on it, what it is building, and whether to reach out now. It is for early-stage investors, platform and talent partners, and angels. It carries one case through every check, wrong turns included, and ends with a written grade of Confirmed, Probable, or Not Forming plus a dated decision. ## What does a "Stealth" title actually tell you? A placeholder title tells you one thing: someone left a role and is not saying what comes next. It does not prove a company exists, that the person is a founder, or that money has moved. Nobody has published how often a "Stealth" title turns into a funded company, so every case starts unproven. The closest public number comes from one flag-based tracker. In one snapshot it had 897 companies flagged, and 243 of them, about 27%, were flagged before their latest round was announced. That tracker flags companies from social-follow signals, not title changes, so treat 27% as a loose reference point, not a base rate for profile titles. Here is what placeholder-title holders look like in Refolk's index of professional profiles. **233 - US profiles in Refolk's index with a Stealth Startup, Stealth Mode, or Building something new title** One public count puts the people listing the Stealth Startup page as their employer at about 38,000, so title search alone misses most of the population. | Metric (Refolk's index) | US | UK | |---|---|---| | Profiles with a placeholder title | 233 | 35 | | Generic employer (Stealth, NewCo, TBA), of 25 sampled | 14 (56%) | 13 (52%) | | Named real employer despite placeholder title, of 25 | 5 (20%) | 4 (16%) | | "Building something new" variants, of 25 | 16 | 17 | | Top metro share, of 25 | SF/Palo Alto 9 (36%) | London 15 (60%) | Three things in that table change how you search: - **The employer field matters more than the title.** Founders tend to put the placeholder in the employer field. Search the employer field, or you'll see only a small fraction of the population. - **One in five US profiles already names the company.** 20% of the US sample showed a named real employer next to a placeholder title. Profiles lag reality, so read the employer field before you open a single registry. - **Seniority labels fail on these profiles.** When I added an Owner/Partner or CXO seniority filter to the US placeholder-title search, it returned 0 profiles. Placeholder-title holders are not classified as founder-level, so judge seniority from the prior role, not the current label. ## The worked case: from 233 profiles to one operator The case begins with triage. I took the 25-profile US sample from Refolk's index, removed the profiles that resolved themselves or were clearly something else, and picked one operator to carry all the way through. The sample counts below are real. Past that point, Refolk does not publish per-case tallies, so the case is a composite. Every fork it hits appears either in the sample or in the record failure modes below. Where I give a count, it is sourced. Where I don't, I describe the fork and leave the number out. ### Triage of the US sample | Pattern in the sample of 25 | Count | What I did | |---|---|---| | Named real employer next to the placeholder | 5 | Resolved from the employer field, no registry work needed | | Title reads "Garden Leave" | 1 | Parked as likely Not Forming | | Title reads "Founding Partner & Co-Head of Investments" | 1 | Routed to the fund check (Form D industry group) | | Employer field reads "Self-employed" | 1 | Low priority, likely consulting | | Generic employer (Stealth, NewCo, TBA) | 14 | The working pool | The first wrong turn came before triage. I tried to narrow the 233 to "founders only" with the seniority filter, got 0, and briefly took that to mean the pool was junk. It wasn't. The labels are simply missing for this population. Seniority has to come from what the person did before. ### The chosen operator and the baseline From the generic-employer pool I picked one profile in the SF/Palo Alto cluster, the largest metro group in the US sample at 9 of 25. The operator was a senior engineer with a long tenure at a large AI research lab. The title read "Building something new" and the employer field read "Stealth Startup". The baseline took about 15 minutes: - **Title-change date** (Day 0 in everything that follows) - **Prior employer and role**, which is my seniority signal in place of the missing label - **Tenure**, because a long tenure ending with a placeholder looks different from a short stint - **Employer field text**, read verbatim Done is one dated line: who, from where, since when. ## Which public records name the people behind a stealth company, and when? Five public records can name a stealth company's people or owner, and each one appears at a different time. Some name people early, some only months later, and one is missing from more than half of venture rounds. Knowing which record should exist at your point in the timeline is what stops you from reading an absence as an answer. | Record | Who it names | When it appears | |---|---|---| | Delaware name search | No officers | At formation | | Delaware annual report | Directors and officers | March 1 of the following year, corporations only | | California Statement of Information (SI-550) | CEO, Secretary, CFO | Within 90 days of foreign qualification | | Form D | Officers, directors, promoters | 15 days after first sale; filed in under half of rounds | | USPTO trademark application | Owner entity | At filing | #### The order in which stealth evidence usually surfaces 1. **People** - Title changes and co-departures from the same employer 2. **Domains** - Certificates logged within minutes to hours, 24 hours at most 3. **State filings** - Delaware formation, then California qualification and SI-550 within 90 days 4. **Trademarks** - Owner entity and goods and services, at filing 5. **Funding** - Form D within 15 days of first sale, if filed at all *People signals tend to arrive first and filings last, so search in the order the evidence appears, not the order that feels most official.* That order is a typical sequence, not a guarantee. Trademarks in particular can be filed at any point. The detail on each record: - **Delaware.** The free name search confirms an entity exists. An online status check costs $10 per entity, or $20 for more detail, including the current franchise tax assessment and filing history. Neither lists officers. Officers first appear in the annual report, which corporations file by March 1. The first report is due the calendar year after formation. Delaware LLCs pay a tax but file no annual report at all. - **California.** A foreign corporation qualifying to do business in California files a Statement of Information within 90 days. It must list all three statutory officers: Chief Executive Officer, Secretary, and Chief Financial Officer. The statement is open to public inspection and must be offered through an online database. Foreign LLCs file the LLC-12 within 90 days of registration approval. Missing the SI-550 deadline carries a $250 penalty, which gives founders some reason to file on time. - **Form D.** Due within 15 calendar days after the first sale in an offering. Item 3 lists executive officers, directors, and promoters, and Item 7 gives the date of first sale. Filing late does not void the Rule 506 exemption, which is one reason so many issuers skip it. - **USPTO.** The TSDR Status tab shows goods and services, filing basis, current owner, attorney and correspondence details, and prosecution history. - **Certificate transparency (CT).** CT is the public log that every publicly trusted TLS certificate is written to. Sources disagree on how fast certificates become searchable, with estimates ranging from 1 to 5 minutes up to 1 to 3 hours. The guaranteed ceiling is the 24-hour Maximum Merge Delay. I have not researched which other states publish officer names at formation. If your operator is based outside California, check that state's foreign-qualification filing rules before assuming it works the same way. > **Watch out:** A Delaware record with no officers is not a finding > > An officerless Delaware record is exactly what a young company looks like, because the first annual report is not due until the March 1 after formation, and LLCs never file one. Log it as "entity exists, officers not yet public" and go to California. ## The resolution procedure, step by step The procedure has seven checks, takes roughly three hours of analyst time, and runs people-first. Each step has a "done" state, so you can stop at a clear point and pick up later. #### Resolving one placeholder profile 1. **Baseline the profile** - Capture the title-change date, prior employer, role, and tenure, reading the employer field before the title. Done is a dated timeline (15 minutes). 2. **Run the co-departure scan** - List people who left the same employer within plus or minus 90 days and now hold placeholder titles or a shared unnamed employer. Done is a shortlist of 0 to 5 candidate co-founders (30 to 45 minutes). 3. **Search Delaware and California registries** - Search candidate entity names in the Delaware name search and California bizfile, checking for foreign qualification and the SI-550 or LLC-12. Done is a file number plus officer names, or a documented absence (30 minutes). 4. **Query certificate transparency and RDAP** - Run crt.sh with %.domain on every candidate domain and pull RDAP registration data. Done is a dated subdomain list (20 minutes). 5. **Check Form D and trademark ownership** - Search EDGAR for a Form D and run a USPTO owner search with the ON: field tag, then open hits in TSDR. Done is related persons or the trademark owner entity (20 minutes). 6. **Trace early hires and repositories** - Find the GitHub organization's creation date, first members, and founding-role job posts. Done is a dated note on who joined (30 minutes). 7. **Grade and set a dated decision** - Grade Confirmed, Probable, or Not Forming on at least two independent records, name every co-founder, and write a reach-now, watch, or drop decision with a re-check date (15 minutes). Practitioners disagree about the order. Social-signal trackers start from follow graphs, not people moves. Trackers also report that people signals arrive weeks before footprint signals, and that matches my experience. If you run steps 3 to 5 before step 2, they often come back empty because you have no names to search. Run the people scan first. **Lookup sheet for one case** ``` Employer field (verbatim): Co-departure scan: People who worked at and now work alongside a former colleague at an unnamed or stealth company Delaware name search: -> file number, formation date, registered agent California bizfile: -> foreign qualification? SI-550 or LLC-12 officers? crt.sh: %. -> subdomain labels, first-seen dates RDAP: -> registration date, registrar, registrant org (redacted?) EDGAR Form D: -> Item 3 related persons, Item 7 first sale, industry group USPTO Trademark Search, owner field (ON:): -> serial numbers TSDR Status tab: -> current owner, goods and services, filing basis GitHub: org -> creation date, first members Job posts: "" founding engineer ``` *Fill the angle-bracket fields from the baseline and the co-departure shortlist; run each line in order and paste the result date beside it.* ## The teardown: running the case through every check Here the operator goes through the seven checks, with each fork shown as it came up. The most useful wrong turns are at steps 3 and 5. One is a name collision that looks like a hit. The other is a product name that doesn't match any legal entity. ### Step 2: the co-departure scan A co-departure scan lists people who left the same employer within plus or minus 90 days of your operator and now show a placeholder or the same unnamed employer. Co-founders tend to leave together or close together, so this is where the team shows up first. Doing this by hand means paging through the old employer's alumni and checking each title and start date. That's the slow part of the whole job. [Refolk](/) lets you ask for the cohort in one plain-English sentence and get back people with their titles, employer fields, and dates. Here is the same scan run against a real employer: Ask me this: `People who worked at Anthropic and now work alongside a former Anthropic colleague at an unnamed or stealth company` - [run the search](https://www.refolk.ai/start?q=People%20who%20worked%20at%20Anthropic%20and%20now%20work%20alongside%20a%20former%20Anthropic%20colleague%20at%20an%20unnamed%20or%20stealth%20company). *Returns former colleagues whose current employer field is blank, generic, or stealth, with dates you can line up against your operator's Day 0.* For the case, the shortlist landed inside the expected band of 0 to 5. The fork came straight away. One candidate looked like the strongest match, with the same team and a close departure date. But their profile showed no unnamed employer, and a few weeks later they went back to a named employer. That is the sabbatical false positive. It's the same pattern as the "Garden Leave" title in the sample, just without the label. I dropped them from the co-founder list and kept the two candidates whose employer fields matched the operator's. ### Step 3: registries, and the same-name LLC A candidate entity name came from one co-founder's personal site, so I searched it in Delaware. The first hit was an LLC with the exact name. Two things were wrong with it: - **Its formation date was before the operator's title change.** A company formed before anyone left is usually someone else's company. - **Its registered agent didn't match** any agent the candidates' other records pointed to. That is the same-name LLC. A name match on its own proves nothing. The next hit was a Delaware corporation formed after Day 0, with no officers listed, which is expected at that stage. The detailed status check would have shown filing history but still no officers, so I went to California bizfile instead. There the corporation showed a foreign qualification and a filed SI-550. Two of the three statutory officers were the operator and one co-departure candidate. That is record one, and it names people. ### Step 4: certificate transparency and RDAP The first crt.sh query on `%.candidate-domain` returned only the apex domain. Empty CT results usually come from wildcard certificates, which cover every subdomain without naming any, or from subdomains hosted on a platform that issues its own certificates. That mechanism is well understood, but nobody has measured how often it happens, and nobody has published how reliably labels like app, staging, or auth show up in CT. I re-queried after the 24-hour merge ceiling and logged the result with dates either way. The RDAP lookup gave a registration date after Day 0, which fits the timeline. The registrant organization was redacted. Since ICANN's Registration Data Policy took effect, registrars may redact the Registrant Organization when the registrant doesn't consent, and some do so by default. A redacted organization doesn't mean the registrant is an individual. A useful mechanism to know: maximum certificate lifetimes are on a published schedule that steps down from 200 days to 100 and then 47. Shorter lifetimes mean hostnames get re-logged more often, so a re-query a few months out has more chances to catch a new subdomain. ### Step 5: Form D, trademarks, and the product-name mismatch EDGAR had no Form D for the entity. I logged that as checked and absent and moved on. More than 50% of a sample of about 40,000 VC rounds filed no Form D. Separate research found that 31.5% of VC-backed firms never file one. Early-stage and high-tech rounds are the least likely to file, which means the stealth companies you care about most are the least likely to leave this record. The trademark check produced the second wrong turn. The landing page on the candidate domain used a product name, and a USPTO search on that product name found nothing owned by the entity. The product name and the legal entity were different. Searching the legal entity's name with the ON: owner field tag found an application in the entity's name for a different word mark, and the TSDR Status tab showed its goods and services. That is record two, independent of California. It names the owner entity, not people, and its goods and services description is the basis for the thesis guess. If the product name later appears on a mark, check the assignments in TSDR to see whether it moves to the entity. ### Step 6: early hires and repositories The GitHub organization under the product name had a creation date after Day 0, and one of its first members was the second co-founder from the SI-550. A founding-engineer job post appeared under the entity name. Neither is independent enough to carry a grade alone, since both come from the founders themselves. They support the thesis guess and show the team is hiring. That shapes the outreach: an offer of recruiting help lands better when the company is hiring. > Start with the people. Filings arrive late, and a missing Form D tells you almost nothing. ## How stealth resolution goes wrong Most bad resolutions come from treating a coincidence as a match or an absence as an answer. No one has published a taxonomy of stealth false positives, so the list below comes from the US sample in Refolk's index and from how each record works. Each entry gives what the mistake looks like and the record that clears it. - **Sabbatical or garden leave read as a founder.** - Looks like: a placeholder title, sometimes literally "Garden Leave", as in the US sample. - Clears it: no co-departures, no registry entry, and a return to a named employer within months. - **Fund formation read as a startup.** - Looks like: a title such as "Founding Partner & Co-Head of Investments", which appeared in the sample. - Clears it: a Form D whose industry group reads Pooled Investment Fund. - **Same-name LLC.** - Looks like: an exact name match in Delaware. - Clears it: a formation date before the title change or a different registered agent means it is not your company. Confirm the real one through California SI-550 officers that match the candidates. - **Delaware record with no officers read as "no team".** - Looks like: an entity with a file number and nothing else. - Clears it: nothing in Delaware will, until the March 1 after formation for corporations, and never for LLCs. Check the California filing instead. - **Missing Form D read as "not funded".** - Looks like: an empty EDGAR search. - Clears it: nothing. More than 50% of rounds never file, so the absence carries no weight either way. - **Empty CT results read as "no product".** - Looks like: only the apex domain in crt.sh. - Clears it: wildcard certificates and hosted-platform subdomains hide labels. Re-query after 24 hours and again after the next renewal cycle. - **Product name treated as the legal entity.** - Looks like: no trademark or registry hit on the brand on the landing page. - Clears it: the USPTO owner field and the TSDR assignment history, searched under the legal entity's name. - **Redacted registrant organization read as "an individual".** - Looks like: a blank or redacted organization field in RDAP. - Clears it: nothing in RDAP. Redaction has been permitted and often defaulted since the Registration Data Policy took effect, so ignore the field and use registration dates. Two more traps sit upstream of the records. One is the seniority filter returning 0, which looks like an empty pool but only means the labels are missing. The other is the named-employer profile: 20% of the US sample already had the answer in the employer field, and running a full registry sweep on those wastes the afternoon. ## Grading: Confirmed, Probable, or Not Forming A grade is a claim about whether a company exists and who is on it, backed by at least two records that don't share a source. People evidence and entity evidence are separate axes, and the grade depends on where the case sits on both. #### People evidence against entity evidence Horizontal axis runs from Operator alone to Co-founders confirmed. Vertical axis runs from No entity records to Entity records name people or owner. | Quadrant | What it means | | --- | --- | | Records, no team match | Suspect a same-name entity; verify officers and formation date | | Confirmed | Name the co-founders and reach now | | Not Forming or too early | Watch with a dated re-check, or drop if positive evidence points elsewhere | | Probable | Watch until the next expected record, or reach now on a warm path | *A grade needs both axes; strong records without matching people usually mean a name collision, and strong people signals without records usually mean too early.* The three grades: - **Confirmed.** Two independent records establish the entity, and at least one of them names a co-founder alongside your operator. Example: a California SI-550 naming the operator and a co-founder, plus a USPTO application owned by the same entity. - **Probable.** A co-departure pattern plus one footprint record, or two footprint records that point to the same entity but name no people. Example: a co-departure match plus a domain registered after Day 0 with app labels in CT. - **Not Forming.** Two positive records point somewhere other than a new company. Examples: a return to a named employer plus no co-departures, or a fund-partner title plus a Form D in the Pooled Investment Fund group. Absences alone never earn this grade. > **Rule:** Two independent records, or no grade > > Every grade cites at least two records that do not derive from each other; a founder's own site and their own GitHub org count as one source, and an absence never counts at all. The case graded Confirmed. The California SI-550 named the operator and one co-founder, the USPTO owner record tied the entity to a product category, and these two records come from separate public sources. The GitHub org and the job post supported the result but didn't carry it. **Stealth resolution memo** ``` STEALTH RESOLUTION: Resolved on: Baseline: title changed ; prior employer , role , tenure Employer field (verbatim): Grade: Confirmed | Probable | Not Forming Record 1: - - pulled Record 2: - - pulled Why independent: Supporting (not counted): Checked and absent:
(absence is not evidence) Legal entity: Product name (if different): Co-founders named: (named by ), (named by ) Thesis guess: - rests on Decision: Reach now | Watch | Drop Re-check on: because ``` *One memo per operator; file it where the deal team can see it, and keep the re-check date in your calendar, not just in the memo.* ## Reach now, watch, or drop: setting the date The decision follows from the grade, and the date follows from the next record that should appear. Confirmed means reach now. Probable means watch until a specific record is due, unless you have a warm path. Not Forming means drop, with one re-check so a slow starter isn't lost for good. Stealth periods vary widely. One stealth-watching newsletter reported that Monaco spent 19 months in stealth while Sponge spent 2, so a single fixed wait doesn't fit every case. The public flag-to-announcement data shows the range: | Company | Raised | Days flagged before announcement | |---|---|---| | Melius | $20M | 186 | | Vinci | $250M | 151 | | ambi | $26M | 106 | | Trajectory | $40M | 76 | | Crosby | $60M | 33 | | Corma | $60M | 0 | Across all 12 raises on that tracker's list, the median gap was 95 days. That number comes from social-follow flags, not title changes. Still, it suggests that a company you resolve today may be about a quarter away from announcing, which is when a first conversation is worth most. Set the re-check date from the mechanism, in this order of preference: 1. **California SI-550 due.** 90 days from foreign qualification, if you saw the qualification but not the statement yet. 2. **Form D window.** 15 days after a likely first sale, if a raise is rumored. Remember it may never be filed. 3. **Delaware annual report.** The March 1 after formation, for a Delaware corporation with no other officer record. 4. **CT re-query.** 24 hours after an empty result, then again after the next certificate renewal. 5. **Fallback.** About 95 days from Day 0, matching the tracker median. When a Probable case moves to Confirmed, Refolk can rerun the co-departure scan for anyone who joined after your first pass. Early hires often show up there before the company page does, and they tell you what the team is building next. #### Before you file the resolution - [ ] The employer field was read verbatim before any registry search. - [ ] The co-departure scan covered plus or minus 90 days from the title-change date. - [ ] Every Delaware name match was checked for formation date and registered agent against the timeline. - [ ] California foreign qualification and the SI-550 or LLC-12 were searched, or the operator's state was confirmed to be elsewhere. - [ ] crt.sh was queried with a %.domain wildcard and re-queried after 24 hours if empty. - [ ] The USPTO owner search used the legal entity name, not only the product name. - [ ] The grade cites two records that do not derive from each other. - [ ] Every absence is listed as checked and absent, and none is counted toward the grade. - [ ] Each named co-founder is tied to the record that names them. - [ ] The decision is reach now, watch, or drop, with a calendar re-check date and the reason for it. ## Frequently asked questions ### How do I find out what a stealth startup is building? Read the goods and services description on any trademark application the entity owns, then cross-check it against subdomain labels in certificate transparency logs and the skills of early hires. The trademark record is the strongest single clue because the owner must describe the product category at filing. Treat the result as a thesis guess, cite the record it rests on, and confirm it in the first conversation rather than presenting it as fact. ### Where do I find the legal company name of a stealth startup? Start from the people, not the name. Run a co-departure scan to find colleagues who left the same employer, then search their names and any candidate entity names in the Delaware name search and California bizfile. A product name on a landing page often differs from the legal entity, so use the USPTO owner field to connect a product mark to the company that filed it. ### Does a missing Form D mean the stealth company has not raised money? No. In an academic sample of about 40,000 venture rounds, more than half filed no Form D, and early-stage and high-tech rounds were the least likely to file. A filed Form D is strong evidence of a raise and names officers and directors, but its absence tells you almost nothing. Log it as checked and absent, and do not let it lower the grade. ### Why does the Delaware record show no officers? Because Delaware does not publish officers at formation. A Delaware corporation's first annual report, which lists officers and directors, is due by March 1 of the year after formation, and Delaware LLCs never file one. For teams working in California, the Statement of Information filed within 90 days of foreign qualification is usually the earliest public list of the CEO, Secretary, and CFO. ### How long should I wait before re-checking a stealth founder graded Probable? Tie the re-check date to the next record that should appear, not to a fixed habit. If the company has qualified in California, the Statement of Information is due within 90 days. If a raise is likely, a Form D is due 15 days after first sale. As a fallback, one public tracker's flagged companies announced raises a median of 95 days after being flagged. ### How do I tell a real stealth founder from someone on garden leave? Look for positive evidence of a company, not just a placeholder title. Garden leave and sabbaticals usually show no co-departures from the old employer, no registry entry, and a return to a named employer within months. Some profiles say Garden Leave outright. If the co-departure scan comes back empty and no records appear after a re-check, grade the case Not Forming and set a drop date. --- *From the Refolk guide library. I revise these guides rather than replacing them, so the current version is always at https://www.refolk.ai/guides/stealth-founder-resolution-teardown*