BS Report
"I shipped 4 products in 5 weeks using Claude Code"
LinkedIn post — Serhii Korniienko, LinkedIn, ~2026-09-18 · 97,149 impressions, 295 reactions, 152 comments, 4 reposts
2026-09-25 · bullshit-detector 0.14.0
sound lesson, hook built on a false premise
Tally: 9 claims extracted, 5 individually source-checked — 1 confirmed, 1 plausible, 2 misleading, 1 false. 1 unverifiable; 1 not checked; 2 not rateable.
Ambiguous: 0 claims dropped before verification
What it says (neutral summary)
The author built four working products in five weeks with Claude Code, got zero stranger signups, and blames never launching any of them. Lesson: AI made building cheap but not distribution. New rule: 30 days, one product, daily posting, weekly numbers.
Load-bearing claims
The ones the thesis dies without.
| # | Claim | Type | Verdict | Evidence |
|---|---|---|---|---|
| 1 | The author never launched any of [the four products, including the bullshit-detector fact-checking skill] [para 4] — "I never launched any of them" | factual | ❌ false | [2 URLs → 2 origins, judged] 2 URLs → 2 origins Show HN "Bullshit Detector - agent skills that fact-check videos and articles" submitted 2026-07-29 by skorniienko: 65 points, 68 comments. Star timestamps: 77 of 148 stars landed by 2026-07-31, 103 by 2026-08-05 - the launch spike is visible. Traffic referrers also show linkedin.com, reddit.com, dev.ua. Steelman: 'launch' as a sustained daily campaign - but Show HN is the canonical developer launch, and 'any of them' falls on one counterexample. The three other products are private repos, so for them the claim may well hold. (source) |
| 2 | Zero strangers signed up for [any of the four products] [para 2] — "Number of strangers who signed up: 0." | factual | 🟠 misleading | The one product named has no signup flow (open-source skill), so 'signups = 0' is true by construction for it and says nothing. 148 GitHub accounts starred it, most arriving via HN/LinkedIn/Google referrers - i.e. strangers. The post puts '0 strangers' and '140 stars' three sentences apart; true only under a definition of 'signed up' that excludes the only product with traction. Signups for the three private products are unverifiable from outside. (source) |
Incidental claims
| # | Claim | Type | Verdict | Evidence |
|---|---|---|---|---|
| 3 | [The author's] fact-checking skill for Claude has 140 GitHub stars [as of the post, ~2026-09-18] [para 3] | factual | ✅ confirmed | Stargazer timestamps: 144 stars before 2026-09-18, 149 on 2026-09-25. '140' is a round-down, not an inflation. (source) |
| 4 | The author shipped 4 products in 5 weeks using Claude Code [para 1] — "I shipped 4 products in 5 weeks using Claude Code." | factual | 🟡 plausible | Repo creation dates (non-fork): bullshit-detector 07-27, notam-collector 08-11, quibble 08-30, numbered-me 09-06, zulu-to 09-10, raim 09-11. With bullshit-detector as one of the four, the 4th repo lands 07-27 → 09-06 = 41 days (~6 wk) at the earliest; 07-27 → 09-11 = 46 days (6.6 wk) if the four are bullshit-detector/quibble/zulu-to/raim. Repo creation is not ship date and which four are meant is unstated, so '5 weeks' is a mild round-down, not a fabrication. 'Using Claude Code' not checked. (source) |
| 5 | All four [of the author's] products work [para 3] — "All four products work." | anecdote | ❓ unverifiable (by construction) | Three of the candidate repos are private. |
| 6 | [The author's only promotion was] sending links to friends and posting a few times on LinkedIn [para 5] — "I sent links to friends, posted few times here." | anecdote | 🟠 misleading | Omits the Show HN (65 points, 68 comments) - the channel that drove roughly half the stars. (source) |
| 7 | The author has 13 years [of experience] as a backend engineer and tech lead [para 3] — "13 years as a backend engineer and tech lead" | factual | ⚪ not checked | |
| 8 | AI made building almost free but did nothing to the cost of distribution [para 6] | opinion | — | Opinion. Ironically, the post's own data supports it better than its stated facts do: the one product that was launched got ~148 strangers; the unlaunched ones got nothing. |
| 9 | [The author will build] no new projects for 30 days, posting daily on one product and sharing numbers weekly [para 7] | prediction | — | Commitment, not checkable yet. No new public or private repo created since 2026-09-11. |
Tally: 9 claims extracted, 5 individually source-checked — 1 confirmed, 1 plausible, 2 misleading, 1 false. 1 unverifiable; 1 not checked; 2 not rateable.
Ambiguous: 0 claims dropped before verification
Unreachable: 1 source — Reddit search API refused the request. Named in the run record.
Hype signals observed
- Big-number / zero contrast as hook: "4 products in 5 weeks" → "Number of strangers who signed up: 0."
- Credential-then-humility beat: "13 years as a backend engineer and tech lead, and I still made the most basic mistake."
- Anecdote generalised to a rule of the industry: "AI made building almost free."
- Engagement-bait close ("I'd like to hear what broke it for you") — it worked: 152 comments.
Incentive analysis
Author's own post; the headline sells the author as a builder and seeds the 30-day build-in-public series. A cleaner "failure" story (zero, never launched) gets more engagement than the messier truth (one launch with real traction, three without).
Bottom line
The numbers that are stated (140 stars, ~5-6 weeks, four builds) are honest or slightly rounded down. The problem is the framing: "I never launched any of them" is false for the one product the post names - it had a Show HN (65 points, 68 comments) that delivered about half its stars - and "0 strangers" only holds because that product has no signup to count. The lesson survives and is actually stronger with the real facts: the one thing launched got ~148 strangers, the three unlaunched got zero. The post chose the more dramatic version over the one its own data supports.
What a hostile reader would hit first
- The Show HN. Anyone who clicks the repo and searches HN finds a 65-point launch a week after creation. "Never launched" dies in one search. Fix: "I launched one, once, then stopped."
- "0 strangers" next to "140 stars." 140 stars are 140 strangers who cared. Fix: say "0 signups on the paid ones" or "0 users of the other three."
- "5 weeks" — repo dates put it nearer six. Minor; nobody will fight over it.
run: 1m49s, searches 7, tools 9, coverage 0, per claim 22s, model claude-opus-5-5, effort low