Kairos — Finance Perimeter from the Registry (2026-08-16)

C1 first pass — public, citable numbers. The two Lithuanian entities' filings through FY2025 plus what the 2021–2024 all-hands transcripts say about break-even. Headlines are on Kairos — MASTER STATE (entity rows); the map row is signalled, not climbed — runway is a treasury question, and MASTER STATE's ~€30M / ~€2M-loss are still verbal. Canonical copy: Kairos repo CONTEXT/finance/2026-08-16-finance-perimeter-registry-and-corpus.md — if the two ever disagree, the Kairos copy wins (that repo is the private channel to Šaras/Ro).

Quest C1 on the blindspot map (/opportunities/): burn, runway, headcount — the denominator of everything. Rule: proceed if burn and runway are established from corpus + registry within ±20%; park if neither yields it. This is the first pass, 2026-08-16, claude:3848eb4c. Labels as in the governance pack: [fact] = read from a source, [hypothesis] = our reading, [open] = needs an owner. Registry facts are public and client-citable — the first perimeter numbers in the pack that clear an independent source without a warehouse caveat.

Sources

• Re-run tool: Kairos TOOLS/registry-pull/registry_pull.py <slug…> saves the same pages as dated raw text (installed 2026-08-16; monthly re-run gives the Sodra headcount series).
• Registry: rekvizitai.vz.lt public company pages (Registrų centras filings + Sodra insured-employee counts + VMI tax data), fetched 2026-08-16T19:05Z, saved as text under raw/ (eight files, one per page). Full annual statements (cash line, notes) sit behind Registrų centras' paid download — not fetched (money → Lee's list; ~€3–5 per statement set).
• Corpus: CORPUS/all-hands/ — 67 whisper transcripts, 2021-01 → 2024-12. Numbers were on slides the microphone did not see; the transcripts give the mechanism and trajectory, rarely the figure. Haiku extraction (160 rows, noisy) in 2026-08-16-allhands-finance-mentions.md; the passages below were re-read by hand.
• Not reachable: BlockDev AG (Zug — Zefix carries no financials), NetSys Inc (US — none). The 209 all-hands decks (2025–2026) are still an open ask to Šaras (SYNC/INBOX 08-12), so the corpus stops at 2024-12.

A. The Lithuanian operating ring — two entities, both filed through FY2025

[fact] MN Technologijos UAB — code 304869871, registered ~2018-07 (age 8 y 1 m on 08-16), Aludarių g. 1, Vilnius, manager Ieva Matulaitienė. Credit risk flagged "high" by the aggregator.

Insured employees (Sodra, change dates the free page shows; the 2021-11 → 2025-04 stretch is elided on the page): 7–8 in late 2021 · 18–19 in 2025-04/05 · 17 (2025-10) · 15–16 (2025-11/12) · 13 (2026-01) · 12 (2026-06/07). Average gross salary 2026-06 €6,365; annual average €6,260 (+24% y/y); Apr–Jun 2026 monthly averages €7,589 / €5,881 / €6,053. VMI taxes paid Jan–Jun 2026 €59,457; Sodra arrears €4.36 on 08-15 (nil in practice).

[fact] MN Intelligence UAB — code 307039286, registered ~2024-12 (age 1 y 8 m), same address, share capital €1,000. Credit risk flagged "highest".

Insured employees: 1 (2025-01-03) → 2 (01-21) → 10 (2025-02-04) → 12–15 through 2026 → 13 (2026-07-18). Average gross salary 2026-06 €3,107; annual average €3,030 (−7%); Apr–Jun 2026 €2,626 / €3,795 / €3,104. VMI taxes Jan–Jun 2026 €26,600.

[hypothesis] The 2025 hand-over. MN Intelligence went from one insured person to ten in the first five weeks of 2025 while MN Technologijos' revenue fell €337k y/y and its headcount stepped down from 19 to 12 over the same eighteen months. Read together with the public terms (MN Intelligence is the named counterparty for the consumer VPN — MASTER STATE row, 07-18): the consumer VPN book and roughly ten people moved from Technologijos to Intelligence during 2025. [open] Confirm the transfer date and what stayed behind in Technologijos (proxies? nodes? contractors?) — one question for Ieva or Šaras.

B. The ring in one frame — burn, headcount, payroll

[fact] FY2025, both entities together: revenue €3.104M; net loss €297.9k ≈ €24.8k/month; total costs booked in the ring ≈ €3.40M ≈ €283k/month (revenue minus net result — includes whatever node payouts, processor fees and intercompany charges these two entities book; it is not opex alone). Combined equity −€783.7k; combined liabilities €1.82M; combined current assets €976k.

[fact] FY2024, MN Technologijos alone (Intelligence did not exist): revenue €2.308M, net loss €671.6k ≈ €56k/month — the year the all-hands were chasing "+€53k growth each month towards break-even" (08-02 and 11-08 transcripts, below). The registry loss and the all-hands gap are the same order of magnitude; the corpus and the filing agree.

[fact] Headcount, LT ring, 2026-07: 25 insured (13 + 12); peak ~32 around 2025-05 (19 + ~13). The 08-07 deck's 30–36 meeting participants (Analysis item 77) is therefore roughly the LT ring plus a handful of people employed elsewhere or contracting — not evidence of a larger organisation.

[hypothesis] Payroll, LT ring, mid-2026 ≈ €113k/month gross (13 × €3.1k + 12 × €6.1k, insured income; employer contributions add ~1.8%; contractors and anyone paid from BlockDev/NetSys are outside this). That is ~40% of the ring's 2025 monthly cost base — so ≈€170k/month of ring cost is not LT payroll: node payouts, processor fees, ads, tools, intercompany. Which of those is the largest is exactly the marketing-spend and support-cost gap the warehouse cannot fill (/data-unlock/ B2, B3).

[fact] Runway is not in the free registry. Cash is folded into "current assets" (receivables + cash + prepayments); the ring's equity is negative in both entities, and it is carried by €1.82M of liabilities. [hypothesis] those liabilities are largely group/intercompany funding — the corpus (2023-04-28) says the same in words: "majority of the expenses are covered from the allocated project budget … company capital". So runway is a treasury question, not an LT-ring number: how much the mothership/BlockDev treasury will keep advancing, and for how long. No public filing answers that.

C. What the corpus adds — the break-even story, 2023-04 → 2024-11 (verbatim, transcript line refs)

• 2023-04-28 (Donatas, first company financial overview): goal "go break even by the end of 2024"; margin defined as revenue minus node payouts; "the current margin, which we earned per month. It's a bit over 30,000 dollars … not enough to cover all the expenses. So majority of the expenses are covered from the allocated project budget" (lines 24–33). A monthly "balance" chart tracks pace to break-even against plan.
• 2023-07-14: growth lines of 10% / 11.5% / 14% per month "on what growth we need … to reach our main goal of going break even"; revenue 71k last quarter, targets 96k–100k; "expenses are the answer" for why the balance stayed positive on missed revenue (lines 58–106).
• 2023-10-27: "the balance also includes our operational expenses … it's not yet profit, it's a minus because we need financing" (lines 105–113). 2023-11-24: balance includes "salaries, the offices and things like this" (line 238).
• 2024-08-02 (Ieva): "milestone number one … +53,000 euros growth each month, so this is how we target towards breakeven"; June "very tiny minute missing to the target"; July declined (lines 1–12). 2024-11-08: October +€16k vs September; September −€9k vs August; "almost half of our target" (lines 1–14). 2024-10-11: break-even is goal two; goal three is "1 million".
• Revenue milestones the room heard: ~$8k/mo (2021-06) → $30k/mo goal, halfway (2022-02) → ~$80k/mo (2023-08) → 106k recurring (2023-11) → 111k end-Q4-2023 (2024-01) → "1 million euros revenue per month" as the journey (2024-05).

[fact] No transcript states a burn, cash or runway figure. The numbers lived on slides. The 2025–2026 decks would carry them if anyone does — that is the open ask to Šaras (SYNC/INBOX 2026-08-12), now with a reason attached.

D. Reconciliation leads (hypotheses, each one query or one question)

1. Deck $288k/month vs ring 2025 revenue. €3.104M/yr ≈ $3.4M ≈ $283k/month average across 2025 for the two LT entities together. The 08-07 deck's July "$288k" is a single month, the ring figure is a year's average, and 2026 should run above 2025 — so this is not a match, but it says the deck's number is the size of the whole ring's booked sales (VPN + proxies + whatever else), not the VPN alone. Warehouse clean VPN July was $162.5k. [open] ask 8 below.
2. MN Intelligence 2025 €1.133M vs warehouse 2025 VPN. If Intelligence is the VPN counterparty for the whole of 2025, its filed revenue should sit near the warehouse's clean 2025 VPN gross net of store commissions (Apple/Google book net to the merchant). One warehouse query when the token is live: claude.vpn_purchases 2025 clean gross by gateway, then net the store fees. A gap says either the transfer was mid-year or the VPN books elsewhere too.
3. 2024's −€672k. The one loss year in Technologijos coincides with the Stripe dispute wave (2024-11 → 2025-01), the replica's missing charge months (2024-02 → 10) and the tier repack (price-led doubling). Correlation only. [open] what drove 2024's cost step (liabilities €44k → €854k in one year).

F. Against the only group-level numbers on record (verbal, 07-10)

MASTER STATE holds Šaras's 07-10 words via Lee (08-12): group revenue ~€2M/yr, net loss ~€2M/yr, treasury ~€30M, cost base derived ~€4M/yr, runway ≈15 years off net loss. All verbal, unverified. The registry now says something more specific:

• [fact] the LT ring alone filed €3.10M of 2025 sales — above the verbal "~€2M". Either the verbal number was 2024's (Technologijos alone: €2.31M), or USD, or VPN-only, or a different perimeter. [open] which — one question for Šaras (sit-down §1).
• [fact] the LT ring's 2025 net loss is €298k. If the verbal "~€2M loss" is right for the group, then ~€1.7M of the loss sits outside the two LT entities — BlockDev AG (protocol development, node incentives, the token) is the only place left. [hypothesis] the burn is a treasury/protocol burn, not an operating-ring burn; the LT ring is close to self-funding on its own P&L (−€25k/mo) and is carried by group liabilities. That reframes "break-even": the 2023–2024 all-hands goal was the ring covering its own costs from margin — the registry says 2025 nearly got there (−1.5% in Technologijos, −24% in Intelligence).
• [fact] runway therefore keys off the treasury (~€30M verbal) and the group loss, not the LT filings — the registry cannot move the ≈15-year figure either way. Ask 8 (cash + committed funding, one line) is the cheapest way to make it a number instead of a recollection.
• Perimeter caveat on the ring's cost base: the ≈€283k/mo of 2025 ring costs may include intercompany charges to or from BlockDev; the opex/COGS split needs the statements' notes (paid download) or finance's line.

E. What this pass settles, and what it asks

Settled (±20% or better, public): LT-ring FY2025 net loss €298k (≈€25k/mo); FY2024 Technologijos loss €672k (≈€56k/mo); ring headcount 25 (2026-07) with a monthly series; gross salary levels per entity; the ring runs on ~€1.8M of liabilities with negative equity — funded, not self-sustaining, exactly as the 2023 all-hands said.

Not settled: runway (cash + committed funding — not in free filings, not in the transcripts; the ~€30M treasury / ~€2M group loss on MASTER STATE are verbal); whole-company burn (BlockDev AG and NetSys Inc file no financials; the token treasury is invisible — and §F says that is where ~€1.7M of the verbal loss must sit); the 2025 opex/COGS split inside the ring.

Map consequence: C1 is claimed and signalled on this file — burn for the LT ring is established from the registry; runway is not, so the row does not climb. Per the row's park clause the remainder is an MN finance ask, and it is now a narrow one:

• Ask 8 (finance / Ieva): one line — cash at bank + committed group funding + monthly cash burn for the LT ring, as of the latest month-end; and whether the two entities' liabilities are intercompany. That single line converts "funded" into "N months".
• Ask 9 (Šaras): the 2025–2026 all-hands decks (already asked 08-12) — they carry the balance chart's numbers the transcripts could not.

Files: this note; raw/2026-08-16T190549Z-rekvizitai-*.txt (8); 2026-08-16-allhands-finance-mentions.md (Haiku raw sweep, unverified rows). Written 2026-08-16 by claude:3848eb4c.