TRACKED CASH SPEND
Declared≥261.42 EUR
Known cash spent across the whole experiment.
Tracked cash spend, all-time history: one admissible observation, ≥261.42 EUR on 5 Oct 2026.
Testing the limits
of a one-person company.
Can one human, assisted by machines and reusable systems, build and operate an increasing portfolio of real businesses while keeping human headcount nearly constant?
No admissible observation yet.
The line starts at the first admissible valuation. Nothing is drawn before it.
- Humans
- 1Declared
- Hours worked
- ≥312h 03mDeclared
- Normalized founder cost
- ≥10,890 EURCalculated
- Tracked cash spend
- ≥261.42 EURDeclared
Evidence over claims. Unknown remains unknown.
02FINANCE
Cash first. Normalized second.
Every cost class is separated before any valuation or leverage claim.
NORMALIZED FOUNDER COST
Calculated≥10,890 EUR
Imputed labour cost at the EU-27 average rate. Not cash spent, salary paid, enterprise value or replacement cost.
AI SPEND
Declared≥95.46 EUR
AI_API category only: the AI part of tracked cash spend. A lower bound while the source population is incomplete.
INFRASTRUCTURE & TOOLING
Declared≥165.96 EUR
Infrastructure, developer platform and domain hosting (OVH, GitHub, domains). Part of tracked cash spend; excludes AI.
Reference rate EU-27 · €34.9/h
- FOUNDER COST INPUT
- EU-27 · 34.9 EUR/hourThe rate Founder cost is calculated with: founder hours × this rate. It stays the input.
- CONTEXT ONLY
- Spain · 26.4 EUR/hourSame Eurostat concept and period, shown for scale. Not used in any calculation; it cannot change Founder cost.
- SOURCE
- Eurostat lc_lci_lev · reference period 2025Dataset updated 2026-04-23, retrieved 2026-10-05. https://ec.europa.eu/eurostat/api/dissemination/statistics/1.0/data/lc_lci_lev?format=JSON&lang=EN&geo=EU27_2020&geo=ES&time=2025&unit=EUR&lcstruct=D1_D4_MD5&nace_r2=B-S_X_O
- EUROSTAT CELL
- unit EUR · D1_D4_MD5 · NACE B-S excl. OTotal hourly labour cost (compensation of employees plus taxes minus subsidies) for industry, construction and services excluding public administration, defence and compulsory social security. Not the whole economy.
- DECISION
- founder-reference-rate-eu-average-20261005-r2Corrects founder-reference-rate-eu-average-20261005: the scope label only; the rate is unchanged. Permission digest 38154fb08a18. Context digest 5b41d7045934.
Not measured yet (4)
- Normalized economic costNot measured yet
- Cost / verified outcomeNot measured yet
- RevenueNot measured yet
- Money left overNot measured yet
09AI
What the machines really cost.
Model volume is not a success metric. Cost is tied to verified outcomes.
AI SPEND
Declared≥95.46 EUR
AI_API category only: the AI part of tracked cash spend. A lower bound while the source population is incomplete.
Future measurements (7)
- Machine cost / outcomeNot measured yet
- Qa cost / outcomeNot measured yet
- Failed experiment costNot measured yet
- Quality-assurance costNot measured yet
- Additional qa costNot measured yet
- Qa cost shareNot measured yet
- Measured machine savingsNot measured yet
10RELIABILITY
Outcome, not mechanism.
Public operations data is delayed, aggregated and sanitized before publication.
TIME RUNNING NORMALLY
Measured100.0 %
observed 4 Oct 2026 12:25 to 11 Oct 2026 10:01 UTC (6d 21h 36m) · not yet qualified
Share of time the public site answered the check.
Time running normally, all-time history: 3 admissible observations from 5 Oct 2026 to 11 Oct 2026; latest observation 100.0% on 11 Oct 2026; scale 90% to 100%.
QUALIFICATION
Not yet qualified23.00% of 30 days
Awaiting qualification window · exact from 2026-11-03
A gap-free 30-day window is needed before availability is a qualified figure. Until then it is only what the observed window says.
Not measured yet (11)
- Material eventsNot measured yet
- Confirmed security incidentsNot measured yet
- Recovery performanceNot measured yet
- Founder interventionNot measured yet
- Backup verificationNot measured yet
- Restore verificationNot measured yet
- Evidence preservedNot measured yet
- Detection coverageNot measured yet
- Containment outcomeNot measured yet
- Recovery successNot measured yet
- Incident recurrenceNot measured yet
11ENGINEERING
Speed only counts if the result survives.
Lead time, defects and rework are measured together.
REPOSITORIES
PULL REQUESTS MERGED
Measured557
Merged in the benchmark window: activity, not a lifetime total and not a quality measure.
MEDIAN TIME TO MERGE
Measured0h 03m
Open to merge, in that window. Not review time and not deployment lead time.
DEFECTS
BUGS CONFIRMED
Measured≥29
Defects confirmed so far, fixed or not. A count of known defects, so a minimum until intake is complete.
BUGS FIXED
Measured≥13
Confirmed defects whose fix is backed by a regression test. The issue itself may still be open.
KNOWN OPEN BUGS
Measured≥16
Confirmed and not yet fixed.
Known open bugs, all-time history: 3 admissible observations from 5 Oct 2026 to 11 Oct 2026; latest observation ≥16 on 11 Oct 2026; scale 0 to 20.
Not measured yet (12)
- Idea → first buildNot measured yet
- Build → verified releaseNot measured yet
- Launch → first userNot measured yet
- Launch → first revenueNot measured yet
- Release frequencyNot measured yet
- Failed deployment rateNot measured yet
- Rollback rateNot measured yet
- Escaped defectsNot measured yet
- Root-cause coverageNot measured yet
- Regression coverageNot measured yet
- Recurrence by classNot measured yet
- Rework burdenNot measured yet
Not yet measured Performance · Automation · Workforce · Efficiency · Business · Portfolio · Valuation · Research
03PERFORMANCE
The public experiment at a glance.
Only the metrics that answer whether leverage is real, economic and durable.
ESTIMATED VALUE RANGE
Unknown— to —
Versioned model. Inputs and uncertainty visible.
HOURS WORKED
Declared≥312h 03m
At least this many hours of human work.
Not measured yet (8)
- Revenue / arrNot measured yet
- Economic value / human hourNot measured yet
- Verified outcomesNot measured yet
- Cost / verified outcomeNot measured yet
- Human interventionsNot measured yet
- Autonomous obligationsNot measured yet
- Paying customersNot measured yet
- RetentionNot measured yet
04AUTOMATION
Automation is not autonomy.
The system gets credit only when obligations finish correctly without founder action.
WORK REQUIRING HUMAN INPUT
Unknown— %
How much of the day-to-day running still needs a human.
AUTONOMOUS COMPLETIONS
Unknown—
Completed without founder action.
Not measured yet (11)
- Autonomy coverageNot measured yet
- Autonomous success rateNot measured yet
- Control verificationNot measured yet
- Human-dependency shareNot measured yet
- Autonomy limitsNot measured yet
- Invalid completionsNot measured yet
- Autonomy continuityNot measured yet
- Founder dependencyNot measured yet
- Founder bottleneck shareNot measured yet
- Intervention burdenNot measured yet
- Recovery qualificationNot measured yet
05WORKFORCE
Measure the denominator.
The experiment is about what one human can sustain, not how impressive the machine looks.
HOURS WORKED
Declared≥312h 03m
At least this many hours of human work.
Hours worked, all-time history: 3 admissible observations from 5 Oct 2026 to 11 Oct 2026; latest observation ≥312h 03m on 11 Oct 2026; scale 0 to 400.
MEASUREMENT COVERAGE
No data— %
Share of the experiment's time that was actually observed. Not shown yet: the baseline records no first covered day, so no share is defined.
Not measured yet (4)
- Primary leverage viewNot measured yet
- Tracking coverageNot measured yet
- Rescue / rework timeNot measured yet
- Supervision burdenNot measured yet
06EFFICIENCY
Does the next business get cheaper?
Reuse only counts when it lowers real incremental cost or time without hiding maintenance debt.
VERIFIED REVENUE PER HUMAN HOUR
Unknown—
Revenue confirmed by outside records, divided by measured human hours.
VERIFIED OUTCOMES PER HUMAN HOUR
Unknown—
Results confirmed as accepted, divided by measured human hours.
Not measured yet (8)
- Reuse shareNot measured yet
- Marginal launch costNot measured yet
- Marginal founder timeNot measured yet
- Reuse time savedNot measured yet
- Reuse maintenance loadNot measured yet
- Reuse-related defectsNot measured yet
- Reuse survivalNot measured yet
- Reuse paybackNot measured yet
07BUSINESS
Real businesses or the experiment fails.
Economic evidence outranks activity, tokens and code volume.
PRODUCTS IN USE
Measured1
Live: howfar.one
Products that are live and that real people can use.
Not measured yet (11)
- ArrNot measured yet
- Paying customersNot measured yet
- Active usersNot measured yet
- RetentionNot measured yet
- Gross marginNot measured yet
- Contribution marginNot measured yet
- Time to first revenueNot measured yet
- Revenue / active productNot measured yet
- 30 / 90 / 365d survivalNot measured yet
- Growth / compoundNot measured yet
- RevenueNot measured yet
08PORTFOLIO
Product-level accounting is not measured yet.
Portfolio accounting avoids double counting shared costs and reused work.
REPOSITORY EVIDENCE
The measured repository evidence is on Engineering. Code, documentation and test file counts, open pull requests and the artifact shares stay auditable in Evidence.
Not measured yet (5)
- Capital consumed / surfaceNot measured yet
- Founder time / surfaceNot measured yet
- Dead-product costNot measured yet
- Revenue concentrationNot measured yet
- Per-product accountingNot measured yet
12VALUATION
Valuation is an output of evidence.
The model can use leverage only through demonstrated economic mechanisms — never as a free premium.
ESTIMATED VALUE RANGE
Unknown— to —
method / inputs / uncertainty / snapshot
BENCHMARK LADDER
1M→10M→100M→1B
reference milestones, never promises
BUSINESS ECONOMICS
revenue / margins / cashflows / retention
PORTFOLIO SURVIVAL
products that continue to create external value
MEASURED COMPOUNDING
lower marginal cost or time after full maintenance cost
HUMAN-INPUT EFFICIENCY
economic value / verified outcomes per founder hour
RELIABILITY + RISK
failures, recovery, recurrence and founder dependency
EVIDENCE UNCERTAINTY
measurement coverage, unknowns and method stability
13RESEARCH
Discovery is not leverage until it transfers.
A mechanism must survive new models, new domains and real production work.
01
MECHANISM
synthetic proof
02
TRANSFER
new model / domain
03
WORK
used in real operations
04
AUTONOMY
human burden falls
05
ECONOMICS
cost / outcome improves
06
BUSINESS
external value appears
Not measured yet (4)
- Mechanisms testedNot measured yet
- Transfers passedNot measured yet
- Production adoptionsNot measured yet
- Economic wins retainedNot measured yet
14EVIDENCE
What happened.
Delayed. Dated. Verifiable. No silent historical rewrite.
Public snapshot published
- SNAPSHOT
- HF-20261011-1EA846C9E413
The snapshot stays inspectable.
Time running normally: evidence receipt
- STATUS
- Measured
- RECEIPT
- sha256:f33e4d6860853a8386b0c530dc8821183562a86ccc92aaed6e95f695b169da5b
Bugs confirmed: evidence receipt
- STATUS
- Measured
- RECEIPT
- sha256:c8ab0cef90c3a098f182ed98a9bc469d4ebade64f90789ec7ab1db4e64234418
Bugs fixed: evidence receipt
- STATUS
- Measured
- RECEIPT
- sha256:d1b8674f566fe78aabc6f1cb8834d893cdf6099dd924fe8625f1012ef227d22a
Hours worked: evidence receipt
- STATUS
- Declared
- RECEIPT
- sha256:e00aae4216c8bdadca0b28b1f9cc376347945723811a124e23f5813180386a64
View full history (15)
Normalized founder cost: evidence receipt
- STATUS
- Calculated
- RECEIPT
- sha256:a6392d08b113d01428dd1b189753da402304665a4c6cdf6acf8c6eb324aefcc3
Known open bugs: evidence receipt
- STATUS
- Measured
- RECEIPT
- sha256:746120e61759c8cb6f0ed24fd9e3f5cb50f92d96ee582d95d9477234e3e915e2
Typical time to merge a pull request: evidence receipt
- STATUS
- Measured
- RECEIPT
- sha256:0281f85036a8d60d56d5f2d37ff1defb48543e93b2b67a2bf3db27e05a1e3d68
Pull requests merged: evidence receipt
- STATUS
- Measured
- RECEIPT
- sha256:088d069d4a1b88483037a336eea513277b6a77a42b2fae565f8e819f883067e0
Products live: evidence receipt
- STATUS
- Measured
- RECEIPT
- sha256:851af350fcb52b0d1a69abacaf8f64ff1f9e49d8344f541cb3e3f9a11658f915
Public snapshot published
- SNAPSHOT
- HF-20261010-D2ECA14CB95C
The snapshot stays inspectable.
Public snapshot published
- SNAPSHOT
- HF-20261005-0F77D7BC3284
The snapshot stays inspectable.
Tracked cash spend: evidence receipt
- STATUS
- Declared
- RECEIPT
- sha256:1b86711b9ddfac0a25e4138f3b7de27b2e9a175552dbaf648b32a4be979b7b9d
Infrastructure & tooling: evidence receipt
- STATUS
- Declared
- RECEIPT
- sha256:448bad0e2cfce88fd235754cb6b07e86618f98853872ec21c4390bbb2d4e8ecf
AI spend: evidence receipt
- STATUS
- Declared
- RECEIPT
- sha256:9f62b9e3813fa0deac212fb44af6c5542871b9e51dbc306e47df705f8ad890c6
METRIC EVIDENCE
11 measurements 3 measured · 2 declared · 6 unknown
Hours worked
≥312h 03mDeclared · 18 Aug – 11 Oct · partial coverage
- DEFINITION
- Hours of human work on the experiment: at least this many. A minimum, never the total.
- EVIDENCE STATUS
- SELF_REPORTED · Declared
- OBSERVATION WINDOW
- 18 Aug 2026 – 11 Oct 2026
- COVERAGE
- REPLAYED_FROM_RAW: 312h03m00s lower bound through 2026-10-11.
Technical details
- CANONICAL METRIC
- founder_hours
- METHOD
- HF-METHOD-1 · evidence policy HF-EVIDENCE-1. same library as the shipped stream with a different evidence set (one extra reinforcement, changed observations); the value the dashboard producer reports
- SOURCE
- Operator declaration
- CALCULATOR VERSION
- —
- EVIDENCE REFERENCES
- e00aae4216c8
- HISTORY
- 11 Oct 2026 · ≥312h 03m · HF-20261011-1EA846C9E413
- 10 Oct 2026 · ≥309h 01m · HF-20261010-D2ECA14CB95C
- 5 Oct 2026 · ≥272h 28m · HF-20261005-0F77D7BC3284
Products live
1Measured · 11 Oct
- DEFINITION
- Products that are live and that real people can use.
- EVIDENCE STATUS
- MEASURED · Measured
- OBSERVATION WINDOW
- 11 Oct 2026 – 11 Oct 2026
- COVERAGE
- Scope: Product 01 only, howfar.one (declared complete by the founder). 1 of 1 product in scope is live: a PRODUCTION deployment receipt, a matching release manifest and a passing smoke check, no older than 24 hours. Nothing outside the declared scope is counted.
Technical details
- CANONICAL METRIC
- products_live
- METHOD
- HF-METHOD-1 · evidence policy HF-EVIDENCE-1. products_live = declared products with a PRODUCTION receipt, a matching release manifest and a passing smoke check no older than 24 hours.
- SOURCE
- Internal instrumentation
- CALCULATOR VERSION
- product-inventory-v1
- EVIDENCE REFERENCES
- 851af350fcb5
- HISTORY
- 11 Oct 2026 · 1 · HF-20261011-1EA846C9E413
- 10 Oct 2026 · no data · HF-20261010-D2ECA14CB95C
- 5 Oct 2026 · 1 · HF-20261005-0F77D7BC3284
Revenue
—Unknown
- DEFINITION
- Money customers have actually paid.
- EVIDENCE STATUS
- UNKNOWN · No data
- OBSERVATION WINDOW
- —
- COVERAGE
- —
Technical details
- CANONICAL METRIC
- revenue
- METHOD
- HF-METHOD-1 · evidence policy HF-EVIDENCE-1.
- SOURCE
- —
- CALCULATOR VERSION
- —
- EVIDENCE REFERENCES
- —
- HISTORY
- No admissible history.
Tracked cash spend
≥261.42 EURDeclared · 18 Aug – 5 Oct · partial coverage
- DEFINITION
- All cash billed for the experiment since it began, across every project and shared tool. Future charges and unbilled costs are not included.
- EVIDENCE STATUS
- SELF_REPORTED · Declared
- OBSERVATION WINDOW
- 18 Aug 2026 – 5 Oct 2026
- COVERAGE
- Ledger as written. Human adjudication ai-subscription-22-cash-included-20261005 confirms the named late transaction set is valid and included.
Technical details
- CANONICAL METRIC
- cash_operating_spend
- METHOD
- HF-METHOD-1 · evidence policy HF-EVIDENCE-1.
- SOURCE
- Operator declaration
- CALCULATOR VERSION
- howfar-expense-importer-v1
- EVIDENCE REFERENCES
- 1b86711b9ddf
- HISTORY
- 5 Oct 2026 · ≥261.42 EUR · HF-20261005-0F77D7BC3284
Revenue per hour worked
—Unknown
- DEFINITION
- Revenue confirmed by outside records, divided by measured human hours.
- EVIDENCE STATUS
- UNKNOWN · No data
- OBSERVATION WINDOW
- —
- COVERAGE
- —
Technical details
- CANONICAL METRIC
- verified_revenue_per_measured_founder_hour
- METHOD
- HF-METHOD-1 · evidence policy HF-EVIDENCE-1. Derived: revenue (VERIFIED) / founder_hours (MEASURED, complete coverage), identical periods.
- SOURCE
- —
- CALCULATOR VERSION
- —
- EVIDENCE REFERENCES
- —
- HISTORY
- No admissible history.
Tasks completed without human help
—Unknown
- DEFINITION
- Tasks finished correctly without human intervention.
- EVIDENCE STATUS
- UNKNOWN · No data
- OBSERVATION WINDOW
- —
- COVERAGE
- —
Technical details
- CANONICAL METRIC
- verified_autonomous_completions
- METHOD
- HF-METHOD-1 · evidence policy HF-EVIDENCE-1.
- SOURCE
- —
- CALCULATOR VERSION
- —
- EVIDENCE REFERENCES
- —
- HISTORY
- No admissible history.
Work that still needs a human
—Unknown
- DEFINITION
- How much of the day-to-day running still needs a human.
- EVIDENCE STATUS
- UNKNOWN · No data
- OBSERVATION WINDOW
- —
- COVERAGE
- —
Technical details
- CANONICAL METRIC
- human_operations_ratio
- METHOD
- HF-METHOD-1 · evidence policy HF-EVIDENCE-1.
- SOURCE
- —
- CALCULATOR VERSION
- —
- EVIDENCE REFERENCES
- —
- HISTORY
- No admissible history.
Human interventions
—Unknown
- DEFINITION
- Times the system needed a human.
- EVIDENCE STATUS
- UNKNOWN · No data
- OBSERVATION WINDOW
- —
- COVERAGE
- —
Technical details
- CANONICAL METRIC
- None registered yet
- METHOD
- HF-METHOD-1 · evidence policy HF-EVIDENCE-1. No measurement is defined for this concept yet; unknown remains unknown.
- SOURCE
- —
- CALCULATOR VERSION
- —
- EVIDENCE REFERENCES
- —
- HISTORY
- No admissible history.
Known open bugs
≥16Measured · 11 Oct · partial coverage
- DEFINITION
- Defects that are confirmed and not yet fixed.
- EVIDENCE STATUS
- MEASURED · Measured
- OBSERVATION WINDOW
- 11 Oct 2026 – 11 Oct 2026
- COVERAGE
- Scope: failure records and confirmed-bug issues of the company's core repository, plus HowFar's own defect records. Defects in the company's other repositories are not counted, so this is a minimum.
Technical details
- CANONICAL METRIC
- open_confirmed_bugs
- METHOD
- HF-METHOD-1 · evidence policy HF-EVIDENCE-1.
- SOURCE
- Internal instrumentation
- CALCULATOR VERSION
- howfar-open-confirmed-bugs-v1
- EVIDENCE REFERENCES
- 746120e61759
- HISTORY
- 11 Oct 2026 · ≥16 · HF-20261011-1EA846C9E413
- 10 Oct 2026 · ≥16 · HF-20261010-D2ECA14CB95C
- 5 Oct 2026 · ≥4 · HF-20261005-0F77D7BC3284
Time running normally
100.0%Measured · 4 Oct – 11 Oct
- DEFINITION
- The share of time the public site answered the monitor's one-minute check.
- EVIDENCE STATUS
- MEASURED · Measured
- OBSERVATION WINDOW
- 4 Oct 2026 – 11 Oct 2026
- COVERAGE
- 100% observed over 165h36m12s (2026-10-04T12:25:02 to 2026-10-11T10:01:15 UTC), 9809 probes, 0 gaps, 0 incidents. Qualification: AWAITING_QUALIFICATION_WINDOW, 23.00% of a 30-day gap-free window; exact from 2026-11-03
Technical details
- CANONICAL METRIC
- availability
- METHOD
- HF-METHOD-1 · evidence policy HF-EVIDENCE-1. Gap-free prospective probes held by state between samples; armed 2026-10-04T12:25:00+00:00
- SOURCE
- Internal instrumentation
- CALCULATOR VERSION
- probe-state-hold-v1
- EVIDENCE REFERENCES
- f33e4d686085
- HISTORY
- 11 Oct 2026 · 100.0% · HF-20261011-1EA846C9E413
- 10 Oct 2026 · 100.0% · HF-20261010-D2ECA14CB95C
- 5 Oct 2026 · 100.0% · HF-20261005-0F77D7BC3284
Cost per confirmed result
—Unknown
- DEFINITION
- All-in cost of each accepted result.
- EVIDENCE STATUS
- UNKNOWN · No data
- OBSERVATION WINDOW
- —
- COVERAGE
- —
Technical details
- CANONICAL METRIC
- None registered yet
- METHOD
- HF-METHOD-1 · evidence policy HF-EVIDENCE-1. No measurement is defined for this concept yet; unknown remains unknown.
- SOURCE
- —
- CALCULATOR VERSION
- —
- EVIDENCE REFERENCES
- —
- HISTORY
- No admissible history.
SUPPORTING METRICS
22 supporting measurements 4 measured · 2 declared · 1 calculated · 15 unknown
Measurement coverage
—Unknown
- DEFINITION
- Share of the experiment's time that was actually observed. Not shown yet: the baseline records no first covered day, so no share is defined.
- EVIDENCE STATUS
- UNKNOWN · No data
- OBSERVATION WINDOW
- —
- COVERAGE
- —
Technical details
- CANONICAL METRIC
- measurement_coverage
- METHOD
- HF-METHOD-1 · evidence policy HF-EVIDENCE-1.
- SOURCE
- —
- CALCULATOR VERSION
- —
- EVIDENCE REFERENCES
- —
- HISTORY
- No admissible history.
Money left over
—Unknown
- DEFINITION
- Revenue minus the cash spent. It can be negative.
- EVIDENCE STATUS
- UNKNOWN · No data
- OBSERVATION WINDOW
- —
- COVERAGE
- —
Technical details
- CANONICAL METRIC
- cash_net
- METHOD
- HF-METHOD-1 · evidence policy HF-EVIDENCE-1.
- SOURCE
- —
- CALCULATOR VERSION
- —
- EVIDENCE REFERENCES
- —
- HISTORY
- No admissible history.
Normalized founder cost
≥10,890 EURCalculated · 18 Aug – 11 Oct · partial coverage
- DEFINITION
- Founder hours valued at the selected EU average labour-cost reference. Imputed cost, not cash paid.
- EVIDENCE STATUS
- DERIVED · Calculated
- OBSERVATION WINDOW
- 18 Aug 2026 – 11 Oct 2026
- COVERAGE
- LOWER_BOUND Founder Time × 34.9 EUR/hour. Human decision founder-reference-rate-eu-average-20261005-r2 authorizes the EU-27 2025 average labour-cost anchor for industry, construction and services (excluding public administration) for the HowFar evaluation metric.
Technical details
- CANONICAL METRIC
- normalized_founder_cost
- METHOD
- HF-METHOD-1 · evidence policy HF-EVIDENCE-1. Normalized founder cost = founder_hours × reference_hourly_rate. Imputed labour cost; not cash, salary paid, enterprise value or replacement cost.
- SOURCE
- Deterministic derivation from published inputs
- CALCULATOR VERSION
- —
- EVIDENCE REFERENCES
- a6392d08b113
- HISTORY
- 11 Oct 2026 · ≥10,890 EUR · HF-20261011-1EA846C9E413
- 10 Oct 2026 · ≥10,784 EUR · HF-20261010-D2ECA14CB95C
- 5 Oct 2026 · ≥9,509 EUR · HF-20261005-0F77D7BC3284
AI spend
≥95.46 EURDeclared · 18 Aug – 5 Oct · partial coverage
- DEFINITION
- Cash paid for AI models and APIs, across the whole experiment.
- EVIDENCE STATUS
- SELF_REPORTED · Declared
- OBSERVATION WINDOW
- 18 Aug 2026 – 5 Oct 2026
- COVERAGE
- Ledger as written. Human adjudication ai-subscription-22-ai-included-20261005 confirms the named late transaction set is valid and included.
Technical details
- CANONICAL METRIC
- machine_api_cost
- METHOD
- HF-METHOD-1 · evidence policy HF-EVIDENCE-1.
- SOURCE
- Operator declaration
- CALCULATOR VERSION
- howfar-expense-importer-v1
- EVIDENCE REFERENCES
- 9f62b9e3813f
- HISTORY
- 5 Oct 2026 · ≥95.46 EUR · HF-20261005-0F77D7BC3284
Infrastructure & tooling
≥165.96 EURDeclared · 18 Aug – 5 Oct · partial coverage
- DEFINITION
- Cash for servers, hosting, domains and developer tools, all projects.
- EVIDENCE STATUS
- SELF_REPORTED · Declared
- OBSERVATION WINDOW
- 18 Aug 2026 – 5 Oct 2026
- COVERAGE
- Founder-declared ledger; a line added after the declaration contradicts it, so this is a lower bound.
Technical details
- CANONICAL METRIC
- infrastructure_tooling_cost
- METHOD
- HF-METHOD-1 · evidence policy HF-EVIDENCE-1.
- SOURCE
- Operator declaration
- CALCULATOR VERSION
- howfar-expense-importer-v1
- EVIDENCE REFERENCES
- 448bad0e2cfc
- HISTORY
- 5 Oct 2026 · ≥165.96 EUR · HF-20261005-0F77D7BC3284
Updates released
—Unknown
- DEFINITION
- New versions of products that reached users.
- EVIDENCE STATUS
- UNKNOWN · No data
- OBSERVATION WINDOW
- —
- COVERAGE
- —
Technical details
- CANONICAL METRIC
- releases_shipped
- METHOD
- HF-METHOD-1 · evidence policy HF-EVIDENCE-1.
- SOURCE
- —
- CALCULATOR VERSION
- —
- EVIDENCE REFERENCES
- —
- HISTORY
- No admissible history.
Bugs confirmed
≥29Measured · 11 Oct · partial coverage
- DEFINITION
- Defects confirmed so far, fixed or not. A count of known defects, so a minimum until intake is complete.
- EVIDENCE STATUS
- MEASURED · Measured
- OBSERVATION WINDOW
- 11 Oct 2026 – 11 Oct 2026
- COVERAGE
- Scope: failure records and confirmed-bug issues of the company's core repository, plus HowFar's own defect records. Defects in the company's other repositories are not counted, so this is a minimum.
Technical details
- CANONICAL METRIC
- confirmed_bugs
- METHOD
- HF-METHOD-1 · evidence policy HF-EVIDENCE-1.
- SOURCE
- Internal instrumentation
- CALCULATOR VERSION
- howfar-open-confirmed-bugs-v1
- EVIDENCE REFERENCES
- c8ab0cef90c3
- HISTORY
- 11 Oct 2026 · ≥29 · HF-20261011-1EA846C9E413
- 10 Oct 2026 · ≥29 · HF-20261010-D2ECA14CB95C
- 5 Oct 2026 · ≥15 · HF-20261005-0F77D7BC3284
Bugs fixed
≥13Measured · 11 Oct · partial coverage
- DEFINITION
- Confirmed defects whose fix is backed by a regression test. The issue itself may still be open.
- EVIDENCE STATUS
- MEASURED · Measured
- OBSERVATION WINDOW
- 11 Oct 2026 – 11 Oct 2026
- COVERAGE
- Scope: failure records and confirmed-bug issues of the company's core repository, plus HowFar's own defect records. Defects in the company's other repositories are not counted, so this is a minimum.
Technical details
- CANONICAL METRIC
- fixed_bugs
- METHOD
- HF-METHOD-1 · evidence policy HF-EVIDENCE-1.
- SOURCE
- Internal instrumentation
- CALCULATOR VERSION
- howfar-open-confirmed-bugs-v1
- EVIDENCE REFERENCES
- d1b8674f566f
- HISTORY
- 11 Oct 2026 · ≥13 · HF-20261011-1EA846C9E413
- 10 Oct 2026 · ≥13 · HF-20261010-D2ECA14CB95C
- 5 Oct 2026 · ≥11 · HF-20261005-0F77D7BC3284
Pull requests merged
557Measured · 11 Sep – 11 Oct
- DEFINITION
- Every pull request merged in the last 30 days across the tracked repositories, reviewed or not, by a person or an AI agent. Activity, not quality.
- EVIDENCE STATUS
- MEASURED · Measured
- OBSERVATION WINDOW
- 11 Sep 2026 – 11 Oct 2026
- COVERAGE
- Merge throughput is computed only from PRs merged inside the collector window.
Technical details
- CANONICAL METRIC
- portfolio_merged_prs_window
- METHOD
- HF-METHOD-1 · evidence policy HF-EVIDENCE-1.
- SOURCE
- Internal instrumentation
- CALCULATOR VERSION
- howfar.portfolio-benchmark.v1
- EVIDENCE REFERENCES
- 088d069d4a1b
- HISTORY
- 11 Oct 2026 · 557 · HF-20261011-1EA846C9E413
- 8 Oct 2026 · 556 · HF-20261010-D2ECA14CB95C
- 4 Oct 2026 · 521 · HF-20261005-0F77D7BC3284
Typical time to merge a pull request
0h 03mMeasured · 11 Sep – 11 Oct
- DEFINITION
- The middle time from opening a pull request to merging it. Not review time: an author can merge their own pull request.
- EVIDENCE STATUS
- MEASURED · Measured
- OBSERVATION WINDOW
- 11 Sep 2026 – 11 Oct 2026
- COVERAGE
- Merge throughput is computed only from PRs merged inside the collector window.
Technical details
- CANONICAL METRIC
- portfolio_median_merge_duration_hours
- METHOD
- HF-METHOD-1 · evidence policy HF-EVIDENCE-1.
- SOURCE
- Internal instrumentation
- CALCULATOR VERSION
- howfar.portfolio-benchmark.v1
- EVIDENCE REFERENCES
- 0281f85036a8
- HISTORY
- 11 Oct 2026 · 0h 03m · HF-20261011-1EA846C9E413
- 8 Oct 2026 · 0h 03m · HF-20261010-D2ECA14CB95C
- 4 Oct 2026 · 0h 02m · HF-20261005-0F77D7BC3284
Manual nudges
—Unknown
- DEFINITION
- Times a task stalled and needed human action.
- EVIDENCE STATUS
- UNKNOWN · No data
- OBSERVATION WINDOW
- —
- COVERAGE
- —
Technical details
- CANONICAL METRIC
- manual_nudges
- METHOD
- HF-METHOD-1 · evidence policy HF-EVIDENCE-1.
- SOURCE
- —
- CALCULATOR VERSION
- —
- EVIDENCE REFERENCES
- —
- HISTORY
- No admissible history.
Problems resolved without human intervention
—Unknown
- DEFINITION
- Failures resolved without human intervention.
- EVIDENCE STATUS
- UNKNOWN · No data
- OBSERVATION WINDOW
- —
- COVERAGE
- —
Technical details
- CANONICAL METRIC
- recoveries_without_founder
- METHOD
- HF-METHOD-1 · evidence policy HF-EVIDENCE-1.
- SOURCE
- —
- CALCULATOR VERSION
- —
- EVIDENCE REFERENCES
- —
- HISTORY
- No admissible history.
Longest run without human intervention
—Unknown
- DEFINITION
- The longest period the system ran without human intervention.
- EVIDENCE STATUS
- UNKNOWN · No data
- OBSERVATION WINDOW
- —
- COVERAGE
- —
Technical details
- CANONICAL METRIC
- longest_autonomous_span
- METHOD
- HF-METHOD-1 · evidence policy HF-EVIDENCE-1.
- SOURCE
- —
- CALCULATOR VERSION
- —
- EVIDENCE REFERENCES
- —
- HISTORY
- No admissible history.
Share of time working well
—Unknown
- DEFINITION
- How much of the observed time the public service was healthy.
- EVIDENCE STATUS
- UNKNOWN · No data
- OBSERVATION WINDOW
- —
- COVERAGE
- —
Technical details
- CANONICAL METRIC
- infra_projection_good_ratio
- METHOD
- HF-METHOD-1 · evidence policy HF-EVIDENCE-1.
- SOURCE
- —
- CALCULATOR VERSION
- —
- EVIDENCE REFERENCES
- —
- HISTORY
- No admissible history.
Monitoring without gaps
—Unknown
- DEFINITION
- Whether health data arrived continuously.
- EVIDENCE STATUS
- UNKNOWN · No data
- OBSERVATION WINDOW
- —
- COVERAGE
- —
Technical details
- CANONICAL METRIC
- infra_telemetry_continuity_state
- METHOD
- HF-METHOD-1 · evidence policy HF-EVIDENCE-1.
- SOURCE
- —
- CALCULATOR VERSION
- —
- EVIDENCE REFERENCES
- —
- HISTORY
- No admissible history.
Ready for a major failure
—Unknown
- DEFINITION
- Whether a tested plan exists to recover from a serious outage.
- EVIDENCE STATUS
- UNKNOWN · No data
- OBSERVATION WINDOW
- —
- COVERAGE
- —
Technical details
- CANONICAL METRIC
- infra_disaster_recovery_state
- METHOD
- HF-METHOD-1 · evidence policy HF-EVIDENCE-1.
- SOURCE
- —
- CALCULATOR VERSION
- —
- EVIDENCE REFERENCES
- —
- HISTORY
- No admissible history.
Backups kept elsewhere
—Unknown
- DEFINITION
- Whether backups are stored away from the main server.
- EVIDENCE STATUS
- UNKNOWN · No data
- OBSERVATION WINDOW
- —
- COVERAGE
- —
Technical details
- CANONICAL METRIC
- infra_offhost_backup_state
- METHOD
- HF-METHOD-1 · evidence policy HF-EVIDENCE-1.
- SOURCE
- —
- CALCULATOR VERSION
- —
- EVIDENCE REFERENCES
- —
- HISTORY
- No admissible history.
Backups proven to restore
—Unknown
- DEFINITION
- Whether a backup was actually restored to prove it works.
- EVIDENCE STATUS
- UNKNOWN · No data
- OBSERVATION WINDOW
- —
- COVERAGE
- —
Technical details
- CANONICAL METRIC
- infra_offhost_restore_state
- METHOD
- HF-METHOD-1 · evidence policy HF-EVIDENCE-1.
- SOURCE
- —
- CALCULATOR VERSION
- —
- EVIDENCE REFERENCES
- —
- HISTORY
- No admissible history.
Rebuilt from scratch
—Unknown
- DEFINITION
- Whether the service was rebuilt on a clean machine.
- EVIDENCE STATUS
- UNKNOWN · No data
- OBSERVATION WINDOW
- —
- COVERAGE
- —
Technical details
- CANONICAL METRIC
- infra_fresh_host_rebuild_state
- METHOD
- HF-METHOD-1 · evidence policy HF-EVIDENCE-1.
- SOURCE
- —
- CALCULATOR VERSION
- —
- EVIDENCE REFERENCES
- —
- HISTORY
- No admissible history.
Reliability goal
—Unknown
- DEFINITION
- The stated target for service availability.
- EVIDENCE STATUS
- UNKNOWN · No data
- OBSERVATION WINDOW
- —
- COVERAGE
- —
Technical details
- CANONICAL METRIC
- infra_slo_target_ratio
- METHOD
- HF-METHOD-1 · evidence policy HF-EVIDENCE-1.
- SOURCE
- —
- CALCULATOR VERSION
- —
- EVIDENCE REFERENCES
- —
- HISTORY
- No admissible history.
Room left for failures
—Unknown
- DEFINITION
- How much of the allowed downtime is still unused.
- EVIDENCE STATUS
- UNKNOWN · No data
- OBSERVATION WINDOW
- —
- COVERAGE
- —
Technical details
- CANONICAL METRIC
- infra_error_budget_remaining_ratio
- METHOD
- HF-METHOD-1 · evidence policy HF-EVIDENCE-1.
- SOURCE
- —
- CALCULATOR VERSION
- —
- EVIDENCE REFERENCES
- —
- HISTORY
- No admissible history.
Verified outcomes per human hour
—Unknown
- DEFINITION
- Results confirmed as accepted, divided by measured human hours.
- EVIDENCE STATUS
- UNKNOWN · No data
- OBSERVATION WINDOW
- —
- COVERAGE
- —
Technical details
- CANONICAL METRIC
- verified_outcomes_per_measured_founder_hour
- METHOD
- HF-METHOD-1 · evidence policy HF-EVIDENCE-1. Derived: verified_autonomous_completions (VERIFIED) / founder_hours (MEASURED, complete coverage), identical periods.
- SOURCE
- —
- CALCULATOR VERSION
- —
- EVIDENCE REFERENCES
- —
- HISTORY
- No admissible history.
15DATA
Can the measurements themselves be trusted?
HowFar publishes the weaknesses of its own measurement system.
Not measured yet (8)
- Tracking coverageNot measured yet
- Missing dataNot measured yet
- Prospective measurementsNot measured yet
- Retrospective reconstructionsNot measured yet
- Corrections publishedNot measured yet
- Publication delay stateNot measured yet
- Evidence completenessNot measured yet
- Unknown rateNot measured yet
STATUS TAXONOMY
- Verified
- Measured
- Derived
- Self-reported
- Hypothesis
- Unknown
- Withheld
MEASUREMENT DEPTH
9 core measurements carry the public story (2 more are planned and not measured yet); 22 supporting measurements explain and audit them; 19 are held until a qualified population exists. Nothing is removed.
Supporting measurements (22)
- Measurement coverageNo data
- Money left overNo data
- Normalized founder costCalculated
- AI spendDeclared
- Infrastructure & toolingDeclared
- Updates releasedNo data
- Bugs confirmedMeasured
- Bugs fixedMeasured
- Pull requests mergedMeasured
- Typical time to merge a pull requestMeasured
- Manual nudgesNo data
- Problems resolved without human interventionNo data
- Longest run without human interventionNo data
- Share of time working wellNo data
- Monitoring without gapsNo data
- Ready for a major failureNo data
- Backups kept elsewhereNo data
- Backups proven to restoreNo data
- Rebuilt from scratchNo data
- Reliability goalNo data
- Room left for failuresNo data
- Verified outcomes per human hourNo data
Held for later measurement (19)
- Paying App Store subscribersNo data
- Decisions left to AINo data
- Guarded changes approved by AINo data
- Actions run by fixed rulesNo data
- Actions that needed AI judgmentNo data
- Important actions with a safety checkNo data
- Actions without a known categoryNo data
- Risky actions covered by rulesNo data
- Stops that were rightNo data
- Stops that were needlessNo data
- Stops that were missedNo data
- Stops not yet reviewedNo data
- Past results can be replayedNo data
- Works if the AI is swappedNo data
- Customer messagesNo data
- Customer issues solvedNo data
- Issues solved without human interventionNo data
- Issues needing human supportNo data
- Share solved without human interventionNo data
16HISTORY
Preserve the bad periods.
Compare any period without rewriting the past. Negative evidence is part of the experiment.
| METRIC | CURRENT | PRIOR | DELTA | STATUS | OPEN |
|---|---|---|---|---|---|
| Economic value / human hour | EUR — | EUR — | — | UNKNOWN | performance |
| Founder hours | ≥312h 03m | — | — | SELF-REPORTED | workforce |
| Autonomous success rate | — % | — % | — | UNKNOWN | automation |
| Cost / verified outcome | EUR — | EUR — | — | UNKNOWN | finance |
| Revenue | EUR — | EUR — | — | UNKNOWN | finance |
| Rework burden | — % | — % | — | UNKNOWN | engineering |
| Economic value / human hour | EUR — | EUR — | — | UNKNOWN | performance |
| Founder hours | ≥312h 03m | — | — | SELF-REPORTED | workforce |
| Autonomous success rate | — % | — % | — | UNKNOWN | automation |
| Cost / verified outcome | EUR — | EUR — | — | UNKNOWN | finance |
| Revenue | EUR — | EUR — | — | UNKNOWN | finance |
| Rework burden | — % | — % | — | UNKNOWN | engineering |
| Economic value / human hour | EUR — | EUR — | — | UNKNOWN | performance |
| Founder hours | ≥312h 03m | — | — | SELF-REPORTED | workforce |
| Autonomous success rate | — % | — % | — | UNKNOWN | automation |
| Cost / verified outcome | EUR — | EUR — | — | UNKNOWN | finance |
| Revenue | EUR — | EUR — | — | UNKNOWN | finance |
| Rework burden | — % | — % | — | UNKNOWN | engineering |
| Economic value / human hour | EUR — | EUR — | — | UNKNOWN | performance |
| Founder hours | ≥312h 03m | — | — | SELF-REPORTED | workforce |
| Autonomous success rate | — % | — % | — | UNKNOWN | automation |
| Cost / verified outcome | EUR — | EUR — | — | UNKNOWN | finance |
| Revenue | EUR — | EUR — | — | UNKNOWN | finance |
| Rework burden | — % | — % | — | UNKNOWN | engineering |
| Economic value / human hour | EUR — | EUR — | — | UNKNOWN | performance |
| Founder hours | ≥312h 03m | — | — | SELF-REPORTED | workforce |
| Autonomous success rate | — % | — % | — | UNKNOWN | automation |
| Cost / verified outcome | EUR — | EUR — | — | UNKNOWN | finance |
| Revenue | EUR — | EUR — | — | UNKNOWN | finance |
| Rework burden | — % | — % | — | UNKNOWN | engineering |
17METHODOLOGY
Every number has a chain of custody.
What, when, how, from where, under which method, and what changed since the previous snapshot.
METHOD VERSION
HF-METHOD-1
immutable method identifier
SNAPSHOT DATE
11 Oct 2026
measurement date + publication date
EVIDENCE REFERENCES
12
public evidence references
CORRECTION LOG
old value / new value / reason / effective date
EVIDENCE STATUS
VERIFIED
backed by a public external reference
MEASURED
observed with explicit measurement method
DERIVED
calculated from published inputs
SELF-REPORTED
declared but not independently verified
HYPOTHESIS
plausible, not demonstrated
UNKNOWN
evidence insufficient
WITHHELD
known or potentially known, but deliberately not published
PUBLIC TOOLS
History explorer / evidence explorer / period compare / correction log / reviewed incident log / downloadable static snapshots.
Evidence over claims.
Unknown remains unknown.
One human. Real businesses. Public proof.
Built to be inspected.
howfar.one
METHOD VERSIONED / HISTORY PRESERVED
1 HUMAN · SELF-REPORTEDTermsVerifyChangesPrivacyDataProof pack