Bluebook (YC Winter 2025): teardown vs XMarts

In one line: Bluebook sells AI agents that sit on top of the ledger an accountant already uses (QuickBooks, NetSuite, Xero, Fortnox) and do the routine close work, with a human approving each step [site] [blog]. Threat: 3 if XMarts picks accounting as its "one type of firm" (W5), because it is the same buyer and problem, funded, live, with a named customer [press] [blog]. Threat 1 if we pick anything else. It is small (4 people) and its public proof is thin [yc].

1. Fact file

2. Their proposition, reverse engineered

3. Positioning

• Category they claim: AI for accounting; "self-driving accounting"; "zero day close" [site] [yc].
• Category a YC partner would put them in: vertical AI agents for accounting (an action layer on top of the ledger) [yc] (inference).
• What they want to be known for: doing the routine accounting work, with the accountant only approving.
• Where they sit on the map: layers L1/L2/L3 = 1/2/2, buyer = sme (accounting firms), moving toward startup finance teams, business.

4. Three layers vs ours

5. Same and different

6. Wedge overlap

7. Whitespace (where they stop)

• Layer gap: no public sign of a firm-wide brain that other work reads, or of results that change the next run [site]. To own it we need a rule-change log tied to checked runs, with rework falling, on one customer workflow.
• Buyer gap: no UK customer, office or tax content on the site; the stack is Nordic (Fortnox) plus US (QuickBooks, ASC) [site] [blog]. They also moved between buyers in under a year (firms, then biotech finance teams), so focus is loose (inference). To own it we need UK firms (for example, small accountancy practices) who say they have the pain, and a UK-specific workflow (VAT, Making Tax Digital). Both are unproven.
• Number gap: every figure is a claim, and the strongest set is an archetype case, not a named customer [blog]. To own it we need one named customer with a before/after number a partner would sign.
• Price gap: no price is public [site]. To own it we need a price a small practice can pay, tied to the result, and 3 to 5 paid pilots or letters of intent.
• Not a gap: speed. They claim go-live in "days" [blog] (UNVERIFIED).

8. For the YC application

• Proven today: Marton's logs show his system already runs agents that do and check work: about 5,600 jobs, each checked by a different AI, with about 160 failure records that each added an automatic check. The inventory lists no accounting workflow, so we have no proof in Bluebook's craft.
• Draft answer to "how are you different from Bluebook?": DRAFT. "Bluebook sells agents to accountants. Our system already runs checked agent work every day on our own business. We have not yet run an accounting workflow, so if we pick accounting we must prove it with a UK firm." Founders say it their own way. If we do not pick accounting, the honest answer is "different buyer".
• Future claim (needs proof): "Our agents run one workflow for a UK accountancy practice and own one number." Needs one customer workflow run for 4+ weeks with a before/after number, plus a named person with the pain.
• What we understand that they don't: a brain that learns from checked results, not from approvals alone. needs proof; Marton's logs show it on his own work (offer feedback lines fell from 44 to 18 as rules were added), not on accounting.
• Copyability (M3): medium. The learning-loop edge is a feature they could add: they already have checklists, approvals, sandboxed code with logs, and Xero support [site] [blog]. Entering the UK needs UK tax and VAT content and new distribution, and they are a 4-person team already split between the Nordics and US-style finance teams [yc] [blog] (inference). Not hard, so M3 can not score above 3 against Bluebook alone.
• Why now (M4): None found in this teardown. The CPA retirement wave is a long trend, not a dated change, and "agents can now do the work" is general AI progress.
• Do not claim: that we are first or alone with agents for accountants. That we know accounting. "Company brain" as our headline. That nobody checks the agents' work (they flag duplicates and check policy [blog]). Any Bluebook number as fact. That Bluebook is dead or pivoted away because the site is down (we do not know why).

9. Ideas worth taking

1. The checklist as the unit of delivery: a recurring job the buyer already owns, each step given to a person or an agent, ending in an approval inbox. The buyer's job becomes "approve" [site].
2. Sell through a network with one framework agreement while each member still decides (Baker Tilly Sweden) [blog]. A cheap way into many small firms of one type.
3. Run the exact maths in code and log every run so it can be re-run and audited [blog]. Fits our "checked work" story.

10. Unknowns

• Why getbluebook.com is down on 2026-10-19, and whether the company has changed name, domain or plan. Check the YC page, their X account, and the EQT portfolio page.
• Real customer count and revenue. YC says 30 firms, EQT said 20+ pilots; neither says paid.
• Price. Ask a buyer who has seen a quote.
• Whether any UK firm uses it. Search for UK customers; look for Xero partner listings.
• What the Notion careers page says (it did not load). It could show the sales model and price band.
• Whether Jonathan Andersson is still with the company (missing from the YC and About pages).

Sources

• getbluebook.com home, archived copy near 2026-04-16 [site]
• About, archived [site]
• Security, archived [site]
• Product: Assistant, archived [site]
• Product: Background agents, archived [site]
• Announcing Bluebook, 13 Feb 2025 [blog]
• Baker Tilly member firms streamline accounting workflows with Bluebook, 25 Mar 2025 [blog]
• How code execution enables accurate AI accounting, 7 Nov 2025 [blog]
• How a biotech startup runs accounting on Bluebook, 20 Nov 2025 [blog]
• YC company page [yc]
• EQT Ventures backs Bluebook, 13 Feb 2025 [press]
• Tech.eu: Bluebook raises $3M, 13 Feb 2025 [press]
• CPA Practice Advisor: Bluebook raises $3M, 13 Feb 2025 [press]
• Techleap: Bluebook raises EUR 2.8M [press] (seen only in search results; not opened)
• Careers page (Notion) [jobs] (did not load)