Business judgment on AI products
Puts several AIs in the roles of judge, counsel, and jury, and records the whole decision so you can replay who influenced whom.
01
Start inside the user's day · Public facts + workflow reasoning · 2026-08-28
Puts several AIs in the roles of judge, counsel, and jury, and records the whole decision so you can replay who influenced whom.
Public materials do not yet show how users complete this job today or what they replace.
The product targets friction in this job, but public user evidence does not yet show the cost, frequency, or consequence of leaving it unsolved.
The trend is that group AI decisions become a compliance problem when you cannot say who swayed whom. The entry is finance approvals and medical consults that must be replayable: sell the trace, not a debate toy.
It promises a simpler way to complete this job: Puts several AIs in the roles of judge, counsel, and jury, and records the whole decision so you can replay who influenced whom. The exact adoption motive and repeat use are not yet verified.
① Whether ProtoLink clears 1,000 stars in three months — whether the example can; pull the framework along; ② Whether the "controllable topology" result gets cited by independent papers or; production case studies; ③ Whether any finance/healthcare company publicly adopts ProtoLink for traceable; age…
Keep watching. It promises a simpler way to complete this job: Puts several AIs in the roles of judge, counsel, and jury, and records the whole decision so you can replay who influenced whom. The exact adoption motive and repeat use are not yet verified.
in a multi-agent system, make "who may communicate with whom" an explicitly configurable topology instead of letting every agent message freely. Controlling the communication structure moves results more than adding agents — that is the transferable conclusion straight from the experiment.
None. This is a research example inside an open-source (MIT) framework: no; pricing, no product, no corporate entity. ① Whether ProtoLink clears 1,000 stars in three months — whether the example can; pull the framework along; ② Whether the "controllable topology" result gets cited by independent papers or; production case studies; ③ Whether any finance/healthcare company publicly adopts ProtoLink for traceable; age…
Puts several AIs in the roles of judge, counsel, and jury, and records the whole decision so you can replay who influenced whom.
It promises a simpler way to complete this job: Puts several AIs in the roles of judge, counsel, and jury, and records the whole decision so you can replay who influenced whom. The exact adoption motive and repeat use are not yet verified.
An English validation note will follow from the public evidence.
The product claims to help users complete: “Puts several AIs in the roles of judge, counsel, and jury, and records the whole decision so you can”. User evidence has not yet verified pain intensity or the cost of doing without it.
The assessment is recorded; an English explanation is pending.
The assessment is recorded; an English explanation is pending.
The assessment is recorded; an English explanation is pending.
02
Market comparison
The Chinese–English market comparison is not complete yet. A conclusion follows only after its coverage and verifiable evidence are recorded.
03
The call and next move come first; the full read retains the evidence and counterevidence.
A replayable multi-agent "liability tribunal" experiment: multiple agents cast as judge, lawyers, witnesses and jurors talk to each other over agent-to-agent (A2A) messages, and the question is how "who may talk to whom" changes a collective verdict. Every run produces an interactive, replayable HTML report.
nMaroulis (GitHub handle), the author of ProtoLink — a Python multi-agent framework built around the A2A protocol, MIT-licensed, positioned as an alternative to chain-centric frameworks like LangChain: each agent is a self-contained runtime entity with identity, capabilities, lifecycle and task-level communication.
ai_courtroom is an example under examples/ in the ProtoLink repo, not a
standalone product.
Read: what the author is really doing is building observable, experimentable tooling for multi-agent communication; the courtroom is just the narrative shell. The actual product being sold is ProtoLink's orchestration and the idea that communication topology can be controlled.
What it deliberately is not: a general intelligence leaderboard, and it makes no causal claims — only reported observations.
Figuring out "do multiple agents get better answers by talking to each other" used to rely on two things:
Judging the final output — treat the multi-agent system as a black box; trust good results, tweak prompts on bad ones, and nobody knows what happened inside.
Reading raw logs — eyeball transcripts and log streams to match "which message caused which change": slow, lossy, and logs only allow after-the-fact review, not tracing the causal chain between messages.
ai_courtroom replaces "debugging multi-agent collaboration by intuition and log reading" with a machine-viewable, frame-by-frame record where every message carries structured state change. It is effectively putting a flight recorder on agent conversations.
None. This is a research example inside an open-source (MIT) framework: no pricing, no product, no corporate entity.
Read: the commercial imagination lives elsewhere — traceability of multi-agent systems is becoming a compliance requirement, and finance and healthcare will eventually demand "every decision step can be replayed." ProtoLink is betting on being the framework that can replay, before the compliance demand arrives.
| Dimension | Call |
|---|---|
| Founder-product fit | High. The author is both framework author and experiment designer; the example serves the framework |
| Product insight | Turns "who talks to whom" — an overlooked variable — into an experiment, closer to the truth than the default "more messages is better" |
| Execution quality | Replay, hash checks, paired ablation, structured action protocols; methodologically stricter than most HN projects |
| Timing | Early. The traceability need is real, but multi-agent systems are not in production at scale yet, so buyers are scarce |
This is the most carefully built demo in the "observability of multi-agent systems" space. Its smartest move is not the courtroom wrapper but treating communication topology as a controlled variable, with a blunt, useful result: what decides collective-decision quality is not message volume but who may talk to whom. Anyone building multi-agent orchestration can use that.
But it is a research example, not a product. No standalone entry point, no users, no business model, and 19 HN points says the breakout is not happening yet. Its value only materializes once ProtoLink gets adopted by real projects.
The transferable rule: making the process replayable builds more trust than making the output better. In any system with multiple cooperating parties — agents or otherwise — solve "can we replay what happened frame by frame" before optimizing results. The repeated warning that "observed change is not causation" is the same methodology.
① Whether ProtoLink clears 1,000 stars in three months — whether the example can pull the framework along ② Whether the "controllable topology" result gets cited by independent papers or production case studies ③ Whether any finance/healthcare company publicly adopts ProtoLink for traceable agent systems
Product logic: in a multi-agent system, make "who may communicate with whom" an explicitly configurable topology instead of letting every agent message freely. Controlling the communication structure moves results more than adding agents — that is the transferable conclusion straight from the experiment.
Pricing structure: none. Not disclosed.
Unproven. The experimental insight is solid and the methodology is rigorous, but this is a framework example rather than a product, with no users and no business model. What is worth keeping is the methodology — replayable, comparable, observations separated from causation. Come back in three months and check whether ProtoLink actually gained traction.
05
Use these searches when the official site is missing or the current link is only a lead.