Research
What must remain usable when the software that acted is replaced? Bulla Labs investigates that question through open software, constructed demonstrations, recorded experiments, and economic research.
Each investigation must earn its own conclusion. Conventional records may suffice; a proposed mechanism may add more work than it removes. The results below distinguish what was observed from what remains open.
Constructed demonstration
What can another actor carry forward?
A coding job across coordinator replacement
Fictional suppliers, supplied code, simulated spending; replacement runs the same prototype.
Accepted fixes remain available; an unconfirmed shipping attempt stays held across coordinator replacement.
Browser and local review use the same project-authored implementation. Reconstruction does not establish authentic history or settle delivery.
Original scheduling example →Research
When is further work worth its full cost?
The economics of unfinished work
Economic research, including a constructed finite decision model.
Second Close makes a later allocation inspectable when new evidence changes the assessment of an unfinished consequence.
No Bulla-specific economic advantage has been established. Model assumptions are not measured enterprise economics; RF003 did not run its main comparison.
Recorded experiment
What happened when a buyer continued from retained evidence?
Routed buyer continuity
Project-operated purchases through OpenRouter under fixed buyer requirements.
A new process continued using retained evidence. Matched signed records produced the same decisions as ActionReceipts.
Public reproduction is partial. Independent interoperability and reduced integration effort remain unmeasured; the controller is research code, not the released package.
24 requests · 11 usable · 12 rejected · 1 unresolved
Available software
What can someone use today?
Bulla
Released open-source software; local receipt tools.
Create and check ActionReceipts, evaluate supplied reliance policies, and reconcile supplied records.
Integrity checks do not establish occurrence, complete observation, or authority for a remedy.
ActionReceipt · bulla==0.49.4
Economics can reject machinery or deployment under the program’s constitutional commitments.
Program essay
Who gets to operate in a machine economy? →
Can commitments, evidence, backing, and dispute procedures be supplied separately without making one intermediary indispensable? The essay develops a research objective, not an experimental result.
Essay sources →Consequence and correction
A consequence names its eligible predicate and authority. Defeated evidence leaves declared dependents recheckable or unresolved rather than silently cleared.
These research mechanisms address who retains evidence, who may decide a dispute, and what could support a remedy. Their implementation status is listed individually.
Research areas
Records and history
Research
Retained history · What did a witness commit to, and can incompatible signed commitments be established?
same-size fork established
Evidence →Constructed demonstration
Recomputes a historical judgment and later correction from retained artifacts.
10,000 review-all → 5,000 required · 0 incomplete cleared
Evidence →Dependency and correction
Research
Correction and recall · Which later judgments require review?
5,000 review · 5,000 no declared path · 0 incomplete cleared
Evidence →Composition and consequence
Recorded experiment
Recorded inference deliveries, retained buyer decisions and process replacement through OpenRouter.
24 requests · 11 usable · 12 rejected · 1 unresolved
Evidence →Constructed demonstration
Combines witnessed history and declared lineage.
2,500 declared paths to recheck
Evidence →Constructed demonstration
Combines transaction records, fault evidence, covenant terms, and consequence authority.
verified authorization · reported Test-ledger attempt · D0
Evidence →Constructed demonstration
Tests whether independently scoped public evidence shares an authenticated event binding.
unbound public evidence → ESCALATE
Evidence →Constructed demonstration
Tests one receiver-owned effect behind one target-side verification and grant-consumption path.
direct call rejected · 1 grant / 1 committed effect
Evidence →Profiles
Foundational and historical work
Res Agentica examines evidence, production, and legitimate authority in a machine economy. Bulla Labs develops tools and applied experiments addressing particular parts of those problems. Explore the intellectual foundations →