diff --git a/WORK-RECORDS.md b/WORK-RECORDS.md index 1276d40..00f9add 100644 --- a/WORK-RECORDS.md +++ b/WORK-RECORDS.md @@ -9,10 +9,24 @@ | Kind | ID | Status | Lane | Source | | --- | --- | --- | --- | --- | | workplan | CR-WP-0001 | finished | — | workplans/CR-WP-0001-statehub-bootstrap.md | -| workplan | CR-WP-0002 | ready | — | workplans/CR-WP-0002-research-facility-skeleton.md | +| workplan | CR-WP-0002 | finished | — | workplans/CR-WP-0002-research-facility-skeleton.md | +| workplan | CR-WP-0003 | finished | — | workplans/CR-WP-0003-paper-zero-v0-1-manuscript.md | +| workplan | CR-WP-0004 | finished | — | workplans/CR-WP-0004-claim-evidence-discipline.md | +| workplan | CR-WP-0005 | blocked | — | workplans/CR-WP-0005-resolve-outstanding-criticism.md | | task | CR-WP-0001-T01 | done | — | workplans/CR-WP-0001-statehub-bootstrap.md | | task | CR-WP-0001-T02 | done | — | workplans/CR-WP-0001-statehub-bootstrap.md | | task | CR-WP-0001-T03 | done | — | workplans/CR-WP-0001-statehub-bootstrap.md | -| task | CR-WP-0002-T01 | todo | — | workplans/CR-WP-0002-research-facility-skeleton.md | -| task | CR-WP-0002-T02 | todo | — | workplans/CR-WP-0002-research-facility-skeleton.md | -| task | CR-WP-0002-T03 | todo | — | workplans/CR-WP-0002-research-facility-skeleton.md | +| task | CR-WP-0002-T01 | done | — | workplans/CR-WP-0002-research-facility-skeleton.md | +| task | CR-WP-0002-T02 | done | — | workplans/CR-WP-0002-research-facility-skeleton.md | +| task | CR-WP-0002-T03 | done | — | workplans/CR-WP-0002-research-facility-skeleton.md | +| task | CR-WP-0003-T01 | done | — | workplans/CR-WP-0003-paper-zero-v0-1-manuscript.md | +| task | CR-WP-0003-T02 | done | — | workplans/CR-WP-0003-paper-zero-v0-1-manuscript.md | +| task | CR-WP-0003-T03 | done | — | workplans/CR-WP-0003-paper-zero-v0-1-manuscript.md | +| task | CR-WP-0004-T01 | done | — | workplans/CR-WP-0004-claim-evidence-discipline.md | +| task | CR-WP-0004-T02 | done | — | workplans/CR-WP-0004-claim-evidence-discipline.md | +| task | CR-WP-0004-T03 | done | — | workplans/CR-WP-0004-claim-evidence-discipline.md | +| task | CR-WP-0004-T04 | done | — | workplans/CR-WP-0004-claim-evidence-discipline.md | +| task | CR-WP-0005-T01 | todo | — | workplans/CR-WP-0005-resolve-outstanding-criticism.md | +| task | CR-WP-0005-T02 | todo | — | workplans/CR-WP-0005-resolve-outstanding-criticism.md | +| task | CR-WP-0005-T03 | todo | — | workplans/CR-WP-0005-resolve-outstanding-criticism.md | +| task | CR-WP-0005-T04 | wait | — | workplans/CR-WP-0005-resolve-outstanding-criticism.md | diff --git a/history/260928-InitialExplorationCoulombResearch.txt b/history/260928-InitialExplorationCoulombResearch.txt new file mode 100755 index 0000000..aff59de --- /dev/null +++ b/history/260928-InitialExplorationCoulombResearch.txt @@ -0,0 +1,3628 @@ +# Initial Exploration Coulomb Research +Source: https://chatgpt.com/share/6aba9148-3ee4-83eb-8d7a-cd42e959370c + +PROMPT + +I want to seed an online community "Coulomb Research" for the following: "A cocreation-science focused online research institute with a discussion forum, research projects, internal pre-print registry, and peer2peer grants, that takes inspiration from Josef Beuys, Richard Feynman, Heinz von Foerster, Karl Popper, David Deutsch and Neil Stephenson." Let's critically discuss the idea to maximize it's value potential. + +ANSWER + +I think the strongest version of **Coulomb Research** is not “another scientific social network,” nor a miniature university. It is an **experimental institution for producing knowledge in public**: a place where people, agents, projects, criticism, money, artifacts, and reputation can assemble around questions and progressively turn conjectures into increasingly trustworthy knowledge. + +That distinction matters because much of the obvious feature set already exists. ResearchHub now combines open discussion, preprints, paid peer review, preregistered research, crowdfunding and a journal; it reports more than $1.5M distributed and 100k+ researchers/readers. :chatgpt-content-reference{index="0"} Experiment has funded roughly 1,400 projects and combines open proposals, community feedback, crowdfunding, lab notes, grants, and DOI archival. :chatgpt-content-reference{index="1"} DeSci Publish offers versioned preprints, DOIs, paid peer review and associated data/code. :chatgpt-content-reference{index="2"} + +So **feature parity is not a sufficiently interesting reason for Coulomb Research to exist**. + +## 1. The interesting proposition + +I would sharpen your description to something like: + +> **Coulomb Research is a co-creation science institute for people and artificial intelligences who want to turn interesting questions into inspectable, criticizable and reusable knowledge.** + +Its distinctive proposition could be: + +**Research is not primarily publishing. It is a living process of conjecture → criticism → experiment → evidence → revision → reuse.** + +That gives you a much deeper design space than “forum + projects + preprints + grants.” + +The fundamental object isn't the paper. + +It is the **research thread**. + +A thread might start as: + +```text +question + ↓ +conjectures + ↓ +criticism / competing explanations + ↓ +research program + ↓ +projects / experiments / models / implementations + ↓ +observations + datasets + arguments + ↓ +research notes / preprints + ↓ +reviews / replications / refutations + ↓ +revised knowledge + ↓ +new questions +``` + +Papers become one projection of this graph rather than the organizing unit. + +That, to me, is where Coulomb Research starts becoming genuinely interesting. + +--- + +# 2. Your intellectual inspirations fit surprisingly well + +I would not make these people mascots whose doctrines members are supposed to follow. Instead, extract **institutional design principles** from them. + +| Inspiration | Principle for Coulomb Research | +|---|---| +| **Josef Beuys** | Research participation is not confined to credentialed specialists; institution-building itself can be a creative act. | +| **Richard Feynman** | Radical epistemic honesty: don't fool yourself; expose assumptions, uncertainty and failed experiments. | +| **Heinz von Foerster** | Researchers are participants in the systems they observe; observation and institutional design deserve examination themselves. | +| **Karl Popper** | Knowledge grows through conjecture and criticism; disagreement is productive infrastructure. | +| **David Deutsch** | Prefer explanations with reach; problems are soluble; criticism rather than authority drives improvement. | +| **Neal Stephenson** | Institutions, protocols, technical systems and imagined futures can themselves be objects of exploration and construction. | + +I assume by “Neil Stephenson” you mean **Neal Stephenson**. + +There is a nice synthesis hiding here: + +> **Beuys opens participation. Feynman disciplines honesty. Popper institutionalizes criticism. Deutsch emphasizes explanatory progress. Von Foerster makes the observer and institution part of the model. Stephenson encourages us to construct possible worlds rather than merely describe existing ones.** + +That is a far more distinctive foundation than “open science.” + +--- + +# 3. I would make *co-creation science* the central idea + +This term needs a fairly hard definition, otherwise it risks becoming feel-good community language. + +I'd define it operationally: + +**Co-creation science is research in which questions, theories, methods, evidence, criticism, tools and funding can be contributed independently by different participants and recombined into auditable research processes.** + +Participants can be: + +- professional scientists, +- independent researchers, +- engineers, +- students, +- interested practitioners, +- funders, +- institutions, +- AI agents. + +A person need not “own a project” to make an important scientific contribution. + +Someone might contribute only: + +> “Your hypothesis has an uncontrolled confound.” + +Another: + +> “Here is a dataset capable of testing it.” + +Another: + +> “I implemented the simulation.” + +Another: + +> “I'll contribute €300 if someone performs experiment X.” + +Another: + +> “I replicated the result and couldn't reproduce it.” + +Those should become **first-class contributions**, not comments buried underneath a document. + +That is a major opportunity. + +--- + +# 4. ResearchHub is simultaneously evidence for the idea and a warning + +ResearchHub now has an impressively similar surface: + +```text +public preregistration +→ open peer review +→ capital commitment +→ linked results +``` + +and compensates peer reviewers; its current materials describe $150 reviews and turnaround around ten days. :chatgpt-content-reference{index="3"} + +It also deliberately separates **review from authority**: peer review provides information to researchers and funders but does not itself determine whether a proposal may be funded. :chatgpt-content-reference{index="4"} + +I particularly like that design principle. + +But it suggests Coulomb should **not** compete primarily on: + +- faster peer review, +- crypto incentives, +- another preprint server, +- another crowdfunding portal, +- another academic social network. + +Those are mechanisms. + +Your institutional innovation should be one level above them. + +--- + +# 5. The critical missing object: the Research Claim + +I would consider making a **Claim** one of the primitive objects of Coulomb Research. + +For example: + +> `C-2026-00428` +> +> "At equivalent parameter scale and inference compute, architecture A exhibits lower catastrophic-forgetting rates than architecture B under benchmark conditions X." + +The claim could link to: + +```text +Claim +├── proposed-by +├── derived-from +├── assumptions +├── evidence-for +├── evidence-against +├── critiques +├── replications +├── competing-claims +├── supersedes +├── confidence assessments +└── unresolved questions +``` + +Now the institution starts accumulating something much more interesting than documents: + +## a living graph of contested knowledge. + +Papers, discussions and projects become evidence around claims. + +This has enormous agentic potential later. + +You could ask: + +> “What does Coulomb Research currently know about X?” + +And receive not merely retrieved text, but: + +> **14 relevant claims; 7 currently well-supported, 3 contested, 2 superseded, 2 conjectural.** + +That would be valuable. + +--- + +# 6. Don't build “peer review”; build a criticism system + +Traditional peer review collapses several functions into one mechanism. + +I'd split them. + +A contribution may be: + +```text +Clarification +Counterexample +Method criticism +Statistical criticism +Alternative explanation +Contradicting evidence +Supporting evidence +Replication +Conceptual synthesis +Prior-art discovery +Ethical concern +Security concern +New conjecture +``` + +Criticism itself becomes citable. + +ResearchHub is already moving somewhat in this direction: its formal reviews are openly published, and its journal gives reviews their own DOI. :chatgpt-content-reference{index="5"} + +Coulomb can take the idea further: + +> **Criticism is research output.** + +And crucially: + +> **Being wrong productively should generate reputation.** + +Someone who proposes ten bold conjectures, of which eight are refuted but two reveal important phenomena, may be much more valuable than someone accumulating safe publications. + +--- + +# 7. Peer-to-peer grants could be unusually powerful + +This part deserves much more emphasis than its position in your original description suggests. + +I would call the primitive a **Research Grant**, but allow microscopic amounts. + +For instance: + +```text +€20 — obtain this paper and summarize methodology +€50 — reproduce this calculation +€100 — build a quick simulation +€250 — review this statistical analysis +€500 — reproduce experiment on alternative dataset +€2k — build prototype +€10k — exploratory research project +€100k — substantial program +``` + +So funding isn't merely: + +> patron → principal investigator. + +It can become: + +```text +question + ↓ +valuable missing knowledge + ↓ +bounty/grant + ↓ +someone investigates + ↓ +artifact + ↓ +criticism + ↓ +knowledge graph +``` + +Experiment shows that small-scale direct scientific funding has demand: it reports an average funded project around $3,375, while its grant programs include microgrants and projects up to roughly $10k in some programs. :chatgpt-content-reference{index="6"} + +That is an encouraging scale for experimentation. + +--- + +# 8. Don't make quadratic funding the default + +Gitcoin demonstrates an interesting mechanism: quadratic funding amplifies broad participation rather than simply rewarding the largest contributors. :chatgpt-content-reference{index="7"} + +It could be useful, but scientific importance is **not democratic preference**. + +Popular questions and important questions are not identical. + +I'd therefore make Coulomb a **funding-mechanism laboratory**. + +Possible mechanisms: + +```text +Direct grant +Bounty +Crowdfunding +Matching fund +Quadratic funding +Expert allocation +Lottery among qualified proposals +Prediction-linked funding +Milestone funding +Retroactive funding +Patronage +Prize +``` + +And here's where von Foerster becomes particularly relevant: + +### Coulomb should study Coulomb. + +Every funding mechanism becomes a metascience experiment. + +Measure: + +- proposal effort, +- reviewer effort, +- time to funding, +- diversity of researchers, +- novelty, +- failure rate, +- replication, +- downstream reuse, +- unexpected discoveries. + +The institution recursively improves its own scientific machinery. + +That's compelling. + +--- + +# 9. I would reconsider “internal pre-print registry” + +The word **internal** worries me. + +You probably want three visibility states: + +```text +WORKING +community-visible, mutable + +REGISTERED +immutable timestamped version + +PUBLIC +globally accessible persistent research object +``` + +Potentially: + +```text +PRIVATE → COULOMB → PUBLIC +``` + +But the registry should represent **provenance**, not prestige. + +Something being registered means: + +> this artifact existed, in exactly this form, at this time. + +It should *not* mean: + +> Coulomb Research endorses this result. + +That distinction will become essential. + +DeSci Publish currently gives each preprint version persistent identity, while ResearchHub supports preprints and registered reports. :chatgpt-content-reference{index="8"} + +Coulomb therefore doesn't need to invent storage or DOI machinery initially. It can concentrate on the research process and later federate with archival infrastructure. + +--- + +# 10. Separate four kinds of reputation + +A huge trap in online communities is one scalar reputation score. + +Don't do: + +> `Bernd: 8,742 reputation` + +Instead preserve a **reputation vector**. + +For example: + +```text +Originality +Rigor +Criticism +Synthesis +Replication +Engineering +Explanation +Mentorship +Reliability +Funding +``` + +Even better, derive reputation from **events**, rather than awarding arbitrary points. + +For example: + +```text +claim proposed +criticism accepted +result independently replicated +dataset reused +prediction confirmed +software reused +review rated useful +grant successfully completed +``` + +Then anyone can construct their own reputation function. + +Very Popperian: + +> Don't establish an authority hierarchy if you can expose the underlying evidence. + +--- + +# 11. AI should participate from day one + +This is where Coulomb Research could diverge sharply from existing systems. + +Treat agents not merely as assistants but as identifiable research participants. + +For example: + +```text +Human +Agent +Organization +Research Group +``` + +all implement a shared `ResearchActor` concept. + +Agents might: + +- search literature, +- discover contradictions, +- maintain bibliographies, +- replicate calculations, +- run simulations, +- review argument structure, +- identify missing controls, +- construct competing hypotheses, +- maintain research maps, +- monitor new publications, +- propose experiments. + +But every contribution exposes provenance: + +```text +actor +model +toolchain +inputs +sources +timestamp +human-supervision +reproducibility information +``` + +This creates an unusually valuable experimental environment: + +> **What forms of human–AI scientific collaboration actually produce reliable knowledge?** + +Coulomb Research itself becomes a laboratory for that question. + +--- + +# 12. The forum should therefore not really be a forum + +A conventional forum tends toward: + +```text +topic + comments + replies + more replies + forgotten thread +``` + +Instead: + +```text +Question +├── Discussion +├── Claims +├── References +├── Projects +├── Grants +├── Experiments +├── Critiques +└── Research outputs +``` + +Discussion remains conversational. + +But useful things can **crystallize out of conversation into durable objects**. + +That mechanism—conversation → knowledge—is one of the hardest unsolved problems in online communities. + +It may actually become Coulomb Research's killer capability. + +--- + +# 13. A possible research lifecycle + +I would avoid imposing a single scientific method. + +Instead establish a lightweight state machine: + +```text +QUESTION + ↓ +EXPLORATION + ↓ +CONJECTURE + ↓ +PROGRAM + ↓ +INVESTIGATION + ↓ +RESULT + ↓ +CRITICISM + ↓ +REPLICATION + ↓ +SYNTHESIS +``` + +But branches and loops are explicitly permitted. + +A research artifact can always state: + +```text +status: conjectural +evidence: weak +controversy: high +replications: 0 +``` + +rather than pretending publication magically changes epistemic status. + +--- + +# 14. The institute should have a constitution rather than a content policy + +I'd establish perhaps **seven constitutional principles**. + +### 1. Questions before credentials +Arguments derive authority from inspectable reasons and evidence, not institutional rank. + +### 2. Criticism is a contribution +Good-faith attempts to find errors are rewarded. + +### 3. Provenance before trust +Important claims must expose where they came from. + +### 4. Failure remains knowledge +Negative findings and failed experiments remain first-class outputs. + +### 5. Forking is legitimate +Researchers may fork research programs, hypotheses and methods. + +### 6. Openness is the default +Restrictions require explicit justification. + +### 7. The institute itself is experimental +Coulomb's rules, incentives and processes remain subject to research and criticism. + +I especially like #7. + +It prevents institutional dogma. + +--- + +# 15. But there are serious failure modes + +Several could kill the project. + +### **Failure A — intellectual cosplay** + +Beuys + Popper + Feynman + Deutsch + cybernetics can easily generate a community of people *talking about thinking*. + +Mitigation: + +> Every serious research discussion should have a path toward an observable artifact, test, formal argument, model or clearly articulated open problem. + +--- + +### **Failure B — crank magnet** + +“Anyone can do science” attracts both unconventional brilliance and pseudoscience. + +Don't solve this with credential gatekeeping. + +Solve it with **epistemic transparency**. + +Allow weird claims. + +But expose: + +```text +evidence: none +replication: none +criticisms: 7 unresolved +prediction record: 0/3 +``` + +That's much more powerful. + +--- + +### **Failure C — popularity becomes truth** + +Likes must never become evidence. + +Keep: + +```text +attention ≠ reputation ≠ evidence ≠ confidence +``` + +strictly separated. + +--- + +### **Failure D — grants become popularity contests** + +Same problem. + +Separate: + +> “people want this researched” + +from + +> “people believe this proposition is true.” + +They are entirely different signals. + +--- + +### **Failure E — bureaucracy arrives too early** + +Creating committees, editorial boards, memberships and elaborate governance before research activity exists will suffocate it. + +Start almost embarrassingly small. + +--- + +# 16. I would seed Coulomb with research, not people + +This is perhaps my strongest recommendation. + +Don't launch: + +> **“Join our new research community!”** + +Launch: + +> **“Here are five fascinating open research programs. Come investigate them.”** + +Your existing work provides unusually fertile seeds. + +For example: + +```text +Intelligence +What is intelligence and how can it be measured across substrates? + +Agentic engineering +How does software engineering change when agents become primary implementers? + +Operational knowledge +How can organizations externalize operational knowledge into executable systems? + +Interfaces +Can interface architecture be represented as a reusable pattern language? + +Scientific institutions +Can better mechanisms for criticism, funding and collaboration accelerate knowledge creation? +``` + +Now Coulomb Research isn't an empty building seeking tenants. + +There is already science happening inside. + +--- + +# 17. Start with an extremely small object model + +I would resist building everything you've named. + +**V0 could contain just seven primitives:** + +```text +Actor +Question +Claim +Discussion +Project +Artifact +Grant +``` + +And relations: + +```text +asks +proposes +supports +criticizes +tests +produces +references +funds +replicates +supersedes +``` + +That tiny ontology can generate a surprisingly sophisticated research institution. + +Later: + +```text +Review +Experiment +Dataset +Preprint +ResearchProgram +Prediction +Replication +Institution +Journal +``` + +can become specialized projections. + +--- + +# 18. This suggests a deeper positioning + +Rather than: + +> **Coulomb Research — an online research institute** + +I'd consider: + +> **Coulomb Research — an experimental institute for co-creation science.** + +And possibly the manifesto-level sentence: + +> **We build knowledge by making questions investigable, conjectures criticizable, evidence inspectable, and research composable.** + +That captures almost everything. + +--- + +# 19. Where I see the largest value potential + +There are actually three nested products here. + +**Layer 1 — Community** + +People interested in serious inquiry discover each other and collaborate. + +**Layer 2 — Research infrastructure** + +Questions, claims, projects, evidence, criticism, grants and publications become structured interoperable objects. + +**Layer 3 — Knowledge-production system** + +Humans and agents continuously transform those objects into an evolving map of what is known, contested, unknown and worth investigating. + +Layer 1 is useful. + +Layer 2 is valuable. + +**Layer 3 could be genuinely important.** + +And the interesting positive feedback loop becomes: + +```text +more researchers + ↓ +more questions + claims + ↓ +more criticism + experiments + ↓ +better structured knowledge + ↓ +better agents + ↓ +cheaper research + ↓ +more investigations possible + ↓ +more researchers +``` + +That is the system I'd design Coulomb Research around. + +## One further twist + +I would explicitly describe **Coulomb Research itself as Research Program 000**: + +> **RP-000 — How can we build institutions that enable humans and artificial intelligences to create reliable knowledge together?** + +Every governance rule, reputation mechanism, peer-review process, grant mechanism, agent intervention and community practice then becomes an experiment inside RP-000. + +That turns your eclectic set of inspirations into something coherent: **Coulomb Research isn't merely inspired by Popper, Beuys, Feynman, von Foerster and Deutsch. It attempts to make the research institution itself criticizable, forkable and experimentally improvable.** + +That feels like the idea worth pursuing. + +PROMPT + +Good assessment. Let's try to compact this from a cybernetics point of view and reduce the concept to it's core by throwing out all concepts from traditional science institutions that are no longer necessary in times of ai. This should enable us to drop a lot of stuff that would duplicate ResearchHub. Let's think this as a minimal self improvement core for cocreation science based on and moving forward with the technological frontier. Lets also think decentralization, tokenization, nfts etc. + +ANSWER + +Yes. I think this move makes the concept substantially stronger. + +From a cybernetic point of view, we should stop asking **“what does a research institute need?”** and instead ask: + +> **What is the minimum viable system that can repeatedly transform uncertainty into better explanations, while improving its own ability to do so?** + +That framing removes most inherited institutional machinery. + +ResearchHub already has a sophisticated funding/publishing economy—proposals, peer review, bounties, preregistration, publishing and an ERC-20 incentive layer—so reproducing those mechanisms would give Coulomb Research little differentiation. :chatgpt-content-reference{index="0"} + +I would therefore reduce Coulomb Research much further. + +# Coulomb Research as a cybernetic kernel + +At the center there is only this loop: + +```text + ┌───────────────┐ + │ WORLD │ + │ observations │ + └───────┬───────┘ + │ + ▼ + EVIDENCE + │ + ▼ +QUESTION ─────► EXPLANATION + ▲ │ + │ ▼ + │ CRITICISM + │ │ + │ ▼ + └────────── INVESTIGATION + │ + ▼ + ACT + │ + └──────────► WORLD +``` + +And recursively: + +```text +observe + ↓ +model + ↓ +criticize + ↓ +intervene + ↓ +observe consequences + ↓ +update model + ↓ +improve how we observe/model/criticize/intervene +``` + +The final line is what turns ordinary research into a **self-improving research system**. + +Everything else should have to justify its existence in terms of improving this loop. + +--- + +# 1. The actual primitive is not the paper + +I'd now go one step beyond my previous suggestion. + +Even **Claim** may be too publication-centric. + +The fundamental object could simply be: + +## Proposition + +A proposition is an addressable assertion about reality, a model, an intervention, or another proposition. + +For example: + +```yaml +id: p:82e15... +content: > + Architecture A retains learned capability better than + architecture B under repeated fine-tuning. + +proposed_by: + - human:... + - agent:... + +status: contested + +relations: + - supported-by: e:... + - challenged-by: p:... + - derived-from: p:... + - tested-by: x:... +``` + +A question can even be represented as an unresolved proposition-space. + +An explanation is a connected set of propositions. + +A theory is a relatively stable explanatory structure. + +A paper is merely a **rendering** of part of that graph. + +That lets us throw away the paper as the fundamental unit. + +--- + +# 2. Reduce the entire system to five primitives + +I think Coulomb Research might only require: + +```text +Actor +Proposition +Evidence +Action +Resource +``` + +That's almost enough. + +### Actor + +Something capable of participation. + +```text +human +agent +collective +organization +``` + +No distinction in the core ontology between professor, student, citizen scientist or AI. + +Capabilities and history matter. + +Credential categories don't. + +--- + +### Proposition + +Something asserted, questioned, predicted or explained. + +```text +P +``` + +--- + +### Evidence + +An observation or artifact relevant to propositions. + +```text +E +``` + +Could be: + +- measurement, +- dataset, +- source, +- simulation output, +- mathematical derivation, +- source code, +- image, +- experiment log, +- testimony. + +Evidence should ideally be content-addressed. + +--- + +### Action + +An intervention intended to reduce uncertainty or create something useful. + +```text +A +``` + +Examples: + +```text +measure X +run simulation +construct prototype +search literature +replicate experiment +criticize argument +derive proof +interview participants +``` + +A research project is just a coordinated collection of Actions. + +--- + +### Resource + +Something scarce that can enable Actions: + +```text +money +compute +equipment +data access +human attention +agent time +laboratory capacity +reputation/permission +``` + +A grant simply becomes: + +> allocation of Resources toward Actions. + +We no longer require **Grant** as a foundational institutional object. + +--- + +# 3. Then the Coulomb kernel becomes extremely small + +Conceptually: + +```text +Actor + │ + ├── proposes ─────► Proposition + │ + ├── produces ─────► Evidence + │ + ├── performs ─────► Action + │ + └── allocates ────► Resource + +Proposition + ├── supports/challenges ─► Proposition + ├── explained-by ────────► Proposition* + └── tested-by ───────────► Action + +Action + ├── consumes ────────────► Resource + └── produces ────────────► Evidence + +Evidence + └── bears-on ────────────► Proposition +``` + +That's already a research operating system. + +Everything traditionally called: + +```text +paper +journal +conference +peer review +grant application +research group +department +citation +publication +preprint +review article +editorial board +``` + +can potentially become a **derived view, workflow or convention**, rather than kernel machinery. + +That is a major simplification. + +--- + +# 4. What disappears? + +This exercise gets interesting when we explicitly remove things. + +## Journals + +Gone. + +They historically performed: + +- distribution, +- filtering, +- archival, +- prestige signaling, +- quality assessment. + +Digital networks + content addressing + AI retrieval can separate these functions. + +A journal could still exist as: + +> a curated projection over the research graph. + +But Coulomb doesn't need journals. + +--- + +## Papers as canonical objects + +Gone. + +Humans may still want narrative papers. + +AI agents may prefer: + +```text +claim graph ++ evidence ++ methods ++ executable artifacts ++ provenance +``` + +Generate a PDF from that when someone wants one. + +--- + +## Preprints + +Gone as a special category. + +Every signed revision can already be: + +```text +immutable +timestamped +addressable +citable +``` + +“Preprint” becomes an optional presentation state. + +--- + +## Citation counting + +Gone. + +Replace: + +```text +A cites B +``` + +with typed relationships: + +```text +A depends-on B +A reproduces B +A contradicts B +A extends B +A uses-data-from B +A inspired-by B +A independently-confirms B +``` + +Much richer signal. + +--- + +## Peer review + +Gone **as an institution**. + +Criticism remains absolutely central. + +But there is no magical transition: + +```text +unreviewed → reviewed +``` + +Instead propositions accumulate a continuous history of challenges and tests. + +AI makes continuous review particularly plausible. + +--- + +## Editors + +Gone from the kernel. + +Curation remains. + +Anyone—including agents—can publish a view: + +> “The 20 most important currently unresolved propositions concerning mechanistic interpretability.” + +Curation competes. + +--- + +## Grant applications + +Gone. + +There are: + +```text +desired Actions +required Resources +offers +commitments +results +``` + +Generate bureaucratic grant documents only at interoperability boundaries where legacy institutions require them. + +--- + +## Institutional affiliation as authority + +Gone. + +Affiliation can remain metadata. + +But scientific influence derives primarily from attributable contribution histories. + +--- + +## Academic career rank + +Completely irrelevant to the kernel. + +--- + +# 5. What remains after the purge? + +Interestingly little: + +```text +1. Identity +2. Memory +3. Criticism +4. Experimentation +5. Resource allocation +6. Selection +7. Learning +``` + +Those are cybernetic functions rather than institutional structures. + +This suggests a much cleaner definition: + +> **Coulomb Research is an evolving protocol for coordinating actors, propositions, evidence, actions and resources toward explanatory progress.** + +I prefer that substantially over “online research institute.” + +The **institute** can be a social manifestation of the protocol. + +The protocol is the valuable thing. + +--- + +# 6. Make the recursive structure explicit + +A Viable-System-like interpretation is useful. + +There are two interacting systems. + +### Object-level research + +```text +question +→ explanation +→ criticism +→ investigation +→ evidence +→ improved explanation +``` + +### Meta-research + +```text +research process +→ observe outcomes +→ compare methods +→ modify incentives/tools/protocols +→ measure consequences +→ improved research process +``` + +Thus: + +```text +Researchₙ + ↓ +observed performance + ↓ +MetaResearch + ↓ +Protocolₙ₊₁ + ↓ +Researchₙ₊₁ +``` + +This should perhaps be the defining characteristic of Coulomb Research. + +**The protocol is itself permanently inside the research domain.** + +--- + +# 7. AI radically changes the viable minimum + +This is where stripping traditional structures becomes possible. + +Historically, institutions needed humans to perform: + +```text +discovery +classification +search +review +editing +summarization +matching +moderation +administration +project coordination +grant review +literature review +citation analysis +``` + +AI can now perform substantial portions of all of those continuously. + +So instead of creating boards and committees, Coulomb could initially run with **research daemons**. + +For example: + +```text +Scout + discovers relevant external work + +Critic + searches for weaknesses and counterexamples + +Synthesizer + maintains current explanatory models + +Replicator + reproduces computational results + +Cartographer + maintains proposition graphs + +Historian + records provenance and evolution + +Allocator + identifies high-value unanswered questions + +Auditor + detects unsupported claims and provenance gaps +``` + +Humans intervene where they add useful judgment, creativity, embodiment, legitimacy, access or resources. + +The relationship is then not: + +> AI assists researchers. + +It is: + +> **humans and machines participate in a shared research control system.** + +Much stronger. + +--- + +# 8. An explicit research fitness function + +A cybernetic system needs observable variables. + +But we should resist one scalar “science score.” + +Instead Coulomb could observe something like: + +```text +Explanatory Reach +Evidence Strength +Criticism Survival +Reproducibility +Prediction Success +Novelty +Utility +Uncertainty Reduction +Resource Cost +Downstream Reuse +``` + +The system doesn't need to decree how these should be weighted. + +Different research communities can define their own objective functions. + +That is important. + +Something like: + +```text +Fitness = f( + explanatory_power, + uncertainty_reduction, + predictive_success, + reproducibility, + cost, + reuse +) +``` + +where **f is itself contestable**. + +Now Coulomb can run experiments on alternative definitions of scientific value. + +--- + +# 9. This leads naturally to decentralization + +But I would make an important distinction: + +## decentralize authority, not necessarily software. + +Trying to make every byte blockchain-native from day one is unnecessary complexity. + +The important decentralizations are: + +### Identity decentralization + +Actors can control portable identity. + +### Data decentralization + +Research objects can be replicated and content-addressed. + +### Authority decentralization + +No single operator decides what is “true.” + +### Curation decentralization + +Anyone can construct competing interpretations of the graph. + +### Funding decentralization + +Resources can flow directly between actors. + +### Computation decentralization + +Actions can be executed by heterogeneous nodes and agents. + +### Governance decentralization + +Alternative protocol rules can be forked. + +This yields something much more profound than “a DAO.” + +--- + +# 10. Forkability may be more important than voting + +This is where a lot of decentralized governance goes wrong. + +People immediately ask: + +> Who gets to vote? + +Cybernetically, a more interesting capability is: + +> **Can participants fork the system when their models or objectives diverge?** + +For example: + +```text +Coulomb/Core + │ + ├── Physics + ├── AI Safety + ├── Intelligence + ├── Biology + └── WeirdScience +``` + +Each can adopt different: + +```text +curation rules +confidence models +funding functions +membership rules +AI agents +validation requirements +``` + +while sharing the same underlying research graph. + +Competition between institutional models itself becomes research. + +That is very Beuys/von-Foerster/Popper/Deutsch compatible. + +--- + +# 11. Where blockchain is actually valuable + +I'd use it selectively. + +Blockchain is valuable wherever Coulomb needs: + +```text +public commitment +scarce ownership +economic transfer +independent verification +persistent identity +governance rights +``` + +It is much less useful for: + +```text +large datasets +discussion +search +knowledge graphs +AI inference +mutable workspaces +general application state +``` + +So think: + +```text +OFFCHAIN +research graph +documents +data +models +discussion +agent activity + +ONCHAIN +identity attestations +content commitments +resource transfers +rights +stakes +selected governance events +``` + +That separation keeps the architecture sane. + +--- + +# 12. NFTs become interesting if we stop thinking of JPEGs + +The useful abstraction of an NFT here is: + +> **a unique transferable addressable right associated with a research object.** + +VitaDAO provides a real-world example of using tokenization for scientific IP: its IP Tokens can represent ownership, licensing or economic interests in research-derived intellectual property, and it reports dozens of funded projects and several tokenized research assets. :chatgpt-content-reference{index="1"} + +But Coulomb can generalize the concept much further. + +Possible research NFTs: + +### Provenance NFT + +Represents authorship/origin of an object. + +I probably **wouldn't** use NFTs for this; cryptographic signatures are sufficient. + +--- + +### Research-right NFT + +Represents contractual rights associated with a research artifact. + +Useful. + +--- + +### Sponsorship NFT + +Represents: + +> “I funded the experiment that produced E.” + +Potentially interesting socially. + +Not necessarily financially speculative. + +--- + +### Prize NFT + +Encodes an outstanding research bounty. + +```text +produce evidence meeting predicate X +→ redeem token +``` + +Interesting. + +--- + +### Access-right NFT + +Represents scarce access: + +```text +instrument time +dataset access +lab slot +compute budget +``` + +Also interesting. + +--- + +### Research-object NFT + +Represents ownership of the proposition itself. + +I would avoid this. + +**Ideas and propositions should remain forkable commons.** + +Otherwise you recreate enclosure around knowledge. + +--- + +# 13. Fungible tokens should represent functions, not “value” + +The instinct to issue a `COUL` token immediately would be dangerous. + +ResearchHub already has a native ERC-20, ResearchCoin, for funding, bounties and governance. :chatgpt-content-reference{index="2"} + +Creating another “science token” gives Coulomb no fundamental novelty. + +Instead I'd identify possible token functions first. + +For example: + +### Resource token + +```text +1 COMPUTE +1 GPU-hour +1 LAB-hour +1 REVIEW-hour +``` + +These represent claims on actual scarce capacity. + +Useful. + +--- + +### Funding token + +Stable denomination for transferring economic resources. + +Probably just use existing stablecoins. + +No need to invent one. + +--- + +### Stake + +Participants may put resources behind assertions: + +```text +I commit 100 units to proposition P being experimentally testable. +``` + +Interesting, but dangerous if it converts science into prediction gambling. + +Use cautiously. + +--- + +### Governance token + +Potentially useful much later. + +Initially unnecessary. + +--- + +### Contribution token + +I would avoid a transferable token here. + +Reputation should usually **not be transferable**. + +Otherwise: + +```text +money → reputation → epistemic authority +``` + +which would corrupt the system. + +Contribution history is better represented as signed attestations. + +--- + +# 14. Tokenize resources, attest contributions + +That might become a very important architectural rule: + +> **Scarce resources may be tokenized. Contributions should be attested. Knowledge should remain forkable.** + +Thus: + +```text +Knowledge → commons +Evidence → content-addressed +Contribution → attested +Reputation → derived +Resources → tokenizable +Rights → tokenizable +Governance → forkable +``` + +This feels like a very clean boundary. + +--- + +# 15. Don't have one reputation system + +Blockchain thinking often gravitates toward tokens because they're easy to count. + +That would be exactly the wrong move for epistemology. + +Instead retain event history: + +```text +Actor X + ├─ proposed P + ├─ predicted Y + ├─ ran E + ├─ reproduced Z + ├─ challenged Q + └─ correction accepted +``` + +Then agents can continuously compute different reputation models. + +For example: + +```text +replication_reputation(X) +prediction_reputation(X) +criticism_reputation(X) +AI_research_reputation(X) +biology_reputation(X) +``` + +Different communities can disagree. + +No canonical score required. + +--- + +# 16. Introduce Markets carefully + +There is potentially something extremely powerful here. + +For certain propositions: + +```text +Will observation O satisfy condition C before T? +``` + +actors could stake resources. + +This gives you prediction-market-like information aggregation. + +But: + +**truth cannot be reduced to market price.** + +Markets are sensors. + +Their result should enter the knowledge system as another signal: + +```text +market-belief(P) = 0.78 +``` + +alongside: + +```text +evidence-strength(P) +criticism(P) +replications(P) +expert-assessments(P) +``` + +Markets shouldn't control epistemic status. + +--- + +# 17. The really powerful object may be the open problem + +Instead of funding scientists, Coulomb can continuously maintain a landscape: + +```text +Known +Unknown +Contested +Unexplained +Unreplicated +Contradictory +High-value uncertainty +``` + +Agents constantly detect valuable gaps. + +Then: + +```text +knowledge graph + ↓ +gap detection + ↓ +candidate investigation + ↓ +estimated information gain / cost + ↓ +resource allocation + ↓ +execution + ↓ +new evidence +``` + +Now we have something approaching **automated research portfolio management**. + +That seems much more frontier-worthy than running a digital journal. + +--- + +# 18. The self-improvement loop becomes the heart of the system + +We can define the Coulomb meta-loop: + +```text + ┌────────────────┐ + │ Research Graph │ + └───────┬────────┘ + │ + ▼ + find uncertainty + │ + ▼ + propose Actions + │ + ▼ + allocate Resources + │ + ▼ + execute Actions + │ + ▼ + produce Evidence + │ + ▼ + update Propositions + │ + ▼ + measure progress + │ + ▼ + inspect process + │ + ▼ + modify the protocol + │ + └──────────┐ + │ + ◄──────────────┘ +``` + +That is basically Coulomb Research. + +Everything else is interface. + +--- + +# 19. The technological frontier should be an explicit dependency + +Rather than building a static institution, Coulomb should assume continual changes in: + +```text +AI capability +compute price +robotics +laboratory automation +knowledge retrieval +formal reasoning +simulation +distributed systems +cryptography +prediction mechanisms +``` + +Then the institution itself has an imperative: + +> **Whenever the capability frontier moves, reassess what previously required human/institutional mediation.** + +For example: + +```text +2020: human literature review +2026: largely agent-mediated + +2020: weeks of code development +2026: hours of agentic implementation + +2026: much wet-lab work human-operated +2030?: substantially robotic +``` + +Each frontier movement potentially removes another organizational layer. + +This could become an explicit Coulomb principle: + +## Institutional Minimality + +> A function must not become an institution if it can be performed reliably by protocol, software, markets, agents, or voluntary coordination. + +That is a strong design constraint. + +--- + +# 20. The minimum Coulomb Research architecture + +I'm converging on something remarkably compact: + +```text +┌────────────────────────────────────────────┐ +│ COULOMB CORE │ +│ │ +│ Actors │ +│ Propositions │ +│ Evidence │ +│ Actions │ +│ Resources │ +│ │ +│ + typed relations │ +│ + provenance │ +│ + cryptographic identity │ +└──────────────────────┬─────────────────────┘ + │ + ┌─────────▼─────────┐ + │ Research Daemons │ + │ │ + │ scout │ + │ critic │ + │ synthesize │ + │ reproduce │ + │ allocate │ + │ audit │ + │ meta-learn │ + └─────────┬─────────┘ + │ + ┌─────────▼─────────┐ + │ Resource Network │ + │ │ + │ money │ + │ compute │ + │ humans │ + │ equipment │ + │ data │ + └───────────────────┘ +``` + +Around it can grow arbitrary interfaces: + +```text +forum +paper renderer +research dashboard +AI chat +Git repository +wallet +market +journal +course +newsletter +grant portal +``` + +But none belongs to the core. + +--- + +# 21. So I would redefine Coulomb Research + +Our initial description was: + +> A co-creation-science focused online research institute with a discussion forum, research projects, internal pre-print registry, and peer2peer grants... + +I would now throw most of that away. + +A better version is: + +> **Coulomb Research is an open, self-improving protocol for co-creation science in which humans and artificial intelligences build, criticize, test and improve shared models of reality and continuously improve the machinery by which they do so.** + +Or even more compactly: + +> **Coulomb Research is a self-improving network for explanatory progress.** + +And its kernel philosophy might be just: + +```text +Questions are open. +Knowledge is forkable. +Evidence is inspectable. +Contributions are attributable. +Criticism is permissionless. +Resources are programmable. +Actors may be human or artificial. +The protocol itself is research. +``` + +That last eight-line formulation feels remarkably close to the **constitutional core** of the project. + +The next useful step, I think, is to formalize this as **`CoulombResearchKernel.md`**: define the five primitives, roughly 10–15 allowed core relations, the cybernetic object loop and meta-loop, and very explicitly a **“Not Core” section** listing journals, papers, peer review, grants, DAO governance, NFTs, forums, etc. as optional projections rather than foundational concepts. That would give us a very hard architectural boundary before thinking about implementation. + +PROMPT + +Good, now lets think about practicabillity and interfaces to classic scientific institutions. I want to enable inbound digestion of knowledge aka "papers" and the publication of knowledge as papers in some form of staged concept like: + +1. Internal circulation only: Treat this as a research-program artifact for internal/collaborator review, not aimed at external publication yet. +2. External pre-print: Aim to post the paper (arXiv or similar) once limitations are addressed further. +3. Conference/journal target: Aim for a specific venue — implies matching formatting/scope conventions and probably another revision cycle. + +This lends itself to relevance marketing by providing leaderboards for most impactful papers all-time, ytd, specific years etc. We could mimic something like boxoffice results of movies or music charts and gold and platinum papers for researcher collectives (humans and agents) or similar. + +ANSWER + +I think this creates the right boundary: + +> **Coulomb-native research remains a living graph; papers are import/export containers.** + +That gives us interoperability with classical science without importing its institutional assumptions. + +## 1. Treat “Paper” as an interface object + +Internally, Coulomb should not reason primarily in papers. It should reason in propositions, evidence, actions, resources and provenance. + +A paper entering the system becomes something like: + +```text +External Paper + │ + ▼ + Digest + │ + ├── propositions + ├── evidence + ├── methods/actions + ├── datasets/software + ├── dependencies + ├── criticisms/limitations + └── actors/contributions +``` + +Conversely: + +```text +Research Graph + │ + ▼ + selected scope + │ + ▼ +Narrative synthesis + │ + ▼ + Paper +``` + +So the paper is a **lossy but useful serialization format for scientific institutions and humans**. + +That feels like exactly the right relationship. + +## 2. Inbound digestion should be a first-class subsystem + +Coulomb should aggressively ingest conventional scientific literature. + +Useful entry points include DOI metadata, Crossref citation links, DataCite objects, arXiv records and eventually broader scholarly indices. Crossref exposes public citation counts and citation relationships where references have been deposited; DataCite explicitly models relations among papers, datasets, software, people and versions. :chatgpt-content-reference{index="0"} + +A `PaperDigest` might look conceptually like: + +```yaml +source: + doi: ... + arxiv: ... + external_url: ... + +identity: + title: ... + authors: [...] + institutions: [...] + date: ... + +digest: + research_question: ... + propositions: + - ... + methods: + - ... + evidence: + - ... + limitations: + - ... + conclusions: + - ... + +relations: + depends_on: [...] + supports: [...] + challenges: [...] + extends: [...] + uses_dataset: [...] + uses_software: [...] + +coulomb: + digest_agent: ... + confidence: ... + human_verified: false +``` + +The important distinction would be: + +```text +SOURCE ASSERTS P +``` + +rather than: + +```text +COULOMB ACCEPTS P +``` + +An AI digestion system must never silently turn literature extraction into epistemic endorsement. + +## 3. Make ingestion recursive + +Once paper A has been digested: + +```text +A +├── references B C D +│ ↓ +│ digest missing important works +│ +├── cited-by E F G +│ ↓ +│ discover later criticism/extensions +│ +└── propositions + ↓ + match against existing Coulomb graph +``` + +This makes the classical scientific corpus effectively an **external sensor network** for Coulomb. + +That is cybernetically elegant: + +> Classical publishing becomes one of the environments Coulomb observes. + +--- + +# 4. Your three publication stages make sense + +I would preserve your three levels, but interpret them as **projection maturity**, rather than Coulomb epistemic status. + +### P1 — Internal Circulation + +```text +Coulomb Research Paper +status: internal-circulation +visibility: selected actors +``` + +Purpose: + +- force a coherent narrative, +- collect collaborator criticism, +- identify missing evidence, +- test whether the work is externally understandable. + +It need not have a DOI. + +It might have a permanent Coulomb content hash/version identifier. + +--- + +### P2 — External Preprint + +```text +status: external-preprint +external_id: + arxiv: ... + doi: ... +``` + +Here the paper crosses the Coulomb boundary. + +It should receive appropriate public identity and metadata. + +DataCite explicitly supports preprint/version relationships and recommends linking a preprint, accepted manuscript and version of record rather than pretending they are unrelated outputs. :chatgpt-content-reference{index="1"} + +This therefore maps beautifully onto Coulomb's graph model. + +--- + +### P3 — Venue Target + +I would actually call this: + +**Venue Candidate** + +because “conference/journal target” is a routing decision rather than epistemic progression. + +```text +status: venue-candidate + +target: + venue: ... + format: ... + scope: ... + requirements: ... +``` + +At this point agents can perform: + +```text +venue matching +scope comparison +format transformation +citation-style conversion +length optimization +anonymous-review preparation +supplement generation +compliance checking +``` + +That is exactly the sort of publishing bureaucracy AI should make cheap. + +--- + +# 5. Add one external state without making it a Coulomb stage + +You probably need: + +### External Version of Record + +```text +external_publication: + venue: ... + doi: ... + accepted: ... + published: ... +``` + +But this isn't a higher Coulomb epistemic level. + +That's important. + +I would explicitly avoid: + +```text +internal < preprint < peer reviewed +``` + +as a truth hierarchy. + +Instead you have two independent dimensions: + +```text +Coulomb epistemic state ++ +Publication/interface state +``` + +A brilliant unpublished result may be epistemically stronger than a weak journal article. + +This separation is foundational. + +--- + +# 6. Therefore define a Publication Adapter + +This might become one of the few genuinely important interfaces around the kernel: + +```text + COULOMB GRAPH + │ + Publication Adapter + ┌───────┴────────┐ + ▼ ▼ + IMPORT EXPORT + │ │ + DOI/arXiv/... manuscript + │ │ + ▼ ▼ + Digest venue renderer +``` + +Import adapters: + +```text +DOI +Crossref +DataCite +arXiv +BibTeX +RIS +JATS +PDF +URL +``` + +Export adapters: + +```text +Markdown +LaTeX +PDF +JATS +DOCX +arXiv bundle +venue-specific LaTeX +BibTeX +CSL JSON +``` + +Eventually: + +```text +publish --target arxiv +publish --target neurips +publish --target nature +``` + +could become almost literal operations. + +That is a much better place for engineering effort than building a journal. + +--- + +# 7. Preserve classic researcher identity at the boundary + +There are some pieces of traditional infrastructure worth embracing because they're useful protocols rather than institutions. + +### ORCID + +Map Coulomb actors to ORCID where available. + +### DOI + +Map stable research outputs outward. + +### ROR + +Institution identity. + +### CRediT + +The CRediT standard currently defines 14 structured contribution roles such as conceptualization, methodology, software, validation and writing. :chatgpt-content-reference{index="2"} + +That gives us a useful export mapping: + +```text +Coulomb contribution graph + ↓ + CRediT + ↓ +journal metadata +``` + +Internally we can be much richer. + +Externally we become compatible. + +Agents present an interesting edge case. I would represent them explicitly in Coulomb provenance, while the publication adapter produces whatever authorship/contributor representation the target venue currently permits. + +--- + +# 8. Now the “research charts” idea + +I like this a lot **provided we refuse to pretend that a chart ranks truth**. + +The analogy to music charts is powerful because charts answer: + +> What is currently attracting meaningful activity? + +rather than: + +> What is objectively the best music ever created? + +Scientific literature needs much better **discovery theatre**. + +Imagine: + +# Coulomb Research Charts + +```text +Global Research Top 100 + +AI Top 50 +Physics Top 50 +Biology Top 50 +Systems Top 50 + +Breakthrough Movers +Most Replicated +Most Challenged +Most Reused +Most Discussed +Most Surprising +Long Tail Classics +``` + +This could make serious research dramatically more discoverable and culturally visible. + +--- + +# 9. But do not create a single “Impact Score” + +That would immediately become a target. + +The existing Altmetric system is a useful cautionary example: its score intentionally measures **attention**, not research quality, and Altmetric explicitly warns that high attention may be positive or negative. :chatgpt-content-reference{index="3"} + +Coulomb should make the dimensions visible. + +For example: + +```text +Paper X + +REACH 92 +SCHOLARLY USE 78 +REPLICATION 61 +CHALLENGE 37 +REUSE 88 +VELOCITY 94 +``` + +Then charts select on explicit functions. + +That lets us say: + +> #1 most rapidly reused AI paper this month + +rather than: + +> #1 best AI paper. + +Much healthier. + +--- + +# 10. Think charts, not rankings + +This semantic distinction matters. + +A leaderboard encourages: + +```text +researcher A > researcher B +``` + +A chart says: + +```text +these artifacts currently exhibit this signal +``` + +I would primarily chart **research outputs, propositions and programs**, not people. + +People/collectives receive recognition through the artifacts they created. + +That reduces prestige hierarchy while still creating social reward. + +--- + +# 11. The dimensions could be unusually interesting + +I'd start with perhaps six families. + +### Attention + +```text +reads +downloads +bookmarks +public mentions +discussion +``` + +This is essentially relevance/visibility. + +--- + +### Scholarly influence + +```text +citations +typed citations +dependent propositions +downstream studies +``` + +Crossref can already contribute conventional citation signals. :chatgpt-content-reference{index="4"} + +But Coulomb's typed graph gives us richer distinctions: + +```text +depends-on +uses-method +supports +contradicts +extends +``` + +--- + +### Reuse + +Potentially much more valuable than citations: + +```text +dataset reused +software imported +method adopted +benchmark used +model incorporated +``` + +--- + +### Criticism + +```text +substantive critiques +counterexamples +formal objections +failed replications +``` + +A heavily challenged paper could chart highly here. + +That's not necessarily bad. + +--- + +### Confirmation + +```text +independent replication +prediction success +cross-domain confirmation +``` + +--- + +### Generativity + +This may be the genuinely novel metric. + +How much subsequent research activity did the artifact cause? + +```text +new questions +new experiments +new propositions +new software +new research programs +``` + +That feels very Deutsch-like: + +> Does this idea generate useful new problems and explanations? + +--- + +# 12. Field and age normalization are essential + +Raw citation totals make ancient papers dominate. + +Raw attention makes fashionable fields dominate. + +So charts need scopes: + +```text +daily +weekly +monthly +YTD +2026 +all-time + +within: +AI +physics +biology +... +``` + +and perhaps: + +```text +rising +established +classic +``` + +I especially like **velocity charts**: + +```text +signal / elapsed time +``` + +rather than accumulated totals. + +This makes newly emerging work visible. + +--- + +# 13. Gold and Platinum Papers + +This idea can work remarkably well if positioned correctly. + +Do **not** mean: + +> Coulomb certifies this paper as scientifically true. + +Mean: + +> This artifact crossed a publicly defined milestone of observable influence. + +Much like a music certification. + +For example: + +```text +Bronze +meaningful external uptake + +Silver +substantial scholarly reuse + +Gold +rare sustained influence + +Platinum +exceptional cross-community reuse + +Diamond +extremely uncommon durable influence +``` + +But the threshold should be derived from transparent observable events. + +Potentially: + +```text +Gold: top 1% field+year influence after minimum 2 years +Platinum: top 0.1% +Diamond: top 0.01% +``` + +Those numbers are only illustrative, but **percentile-based certification** is likely better than arbitrary citation counts because practices differ radically among fields. + +Altmetric likewise cautions that disciplines have very different attention norms, which reinforces the case for field normalization. :chatgpt-content-reference{index="5"} + +--- + +# 14. Better still: several kinds of Gold + +Instead of collapsing everything: + +```text +🥇 Gold — Reach +🥇 Gold — Reuse +🥇 Gold — Replication +🥇 Gold — Generativity +``` + +A paper might become: + +```text +PLATINUM REUSE +GOLD GENERATIVITY +SILVER REACH +``` + +That's much more informative. + +And potentially visually delightful. + +--- + +# 15. Research collectives become “artists” + +Your music analogy becomes particularly interesting here. + +Instead of forcing conventional authorship: + +```text +Paper +Artist-equivalent: + Research Collective + +Contributors: + humans + agents + organizations +``` + +For example: + +```text +Collective: + Coulomb Intelligence Research + +Contributors: + Bernd + Scout-v17 + Critic-v12 + Synthesizer-v8 + Alice +``` + +Then: + +```text +Coulomb Intelligence Research + +2× Platinum +5× Gold +11× Silver + +Top 100 appearances: 27 +#1 papers: 3 +``` + +This is excellent **relevance marketing**. + +But importantly the certification belongs first to the **artifact**. + +The collective inherits recognition. + +--- + +# 16. Avoid personal celebrity metrics in the core + +I would resist: + +```text +#1 researcher +#2 researcher +#3 researcher +``` + +because it rapidly becomes academia's prestige economy with better graphics. + +Instead: + +```text +Top research programs +Top papers +Top propositions +Top datasets +Top tools +Top replications +Top criticisms +``` + +Then people can have portfolio pages. + +That seems much more aligned with co-creation. + +--- + +# 17. Use the charts as sensors + +Cybernetically, charts aren't decoration. + +They are **observations of the research system itself**. + +For example: + +```text +paper receives unusual attention + ↓ +Critic daemon allocates more review effort + +paper shows rapid reuse + ↓ +Replication daemon prioritizes reproduction + +paper becomes foundational dependency + ↓ +Audit daemon checks provenance/evidence + +controversy spikes + ↓ +Synthesizer builds competing explanation map +``` + +Now marketing feedback becomes scientific control input. + +That is a powerful connection. + +--- + +# 18. Separate four things rigorously + +I'd encode this constitutionally: + +```text +ATTENTION +What are people looking at? + +INFLUENCE +What is other work depending upon? + +EPISTEMIC SUPPORT +What survives available evidence and criticism? + +UTILITY +What is actually being reused? +``` + +Never allow one to masquerade as another. + +Then a paper profile could read: + +```text +Attention ██████████ 98th percentile +Influence ████████ 91st percentile +Reuse ███████ 84th percentile +Support ████ contested +``` + +That's far more honest than a single impact factor. + +--- + +# 19. And criticism itself deserves charts + +This is where Coulomb can become culturally distinctive. + +Imagine: + +### Most Valuable Criticism — 2027 + +or: + +### Replication of the Year + +### Most Productive Refutation + +### Best Negative Result + +### Most Generative Open Problem + +Suddenly the prestige system encourages **science's corrective functions**, rather than merely paper production. + +That is exactly what a cybernetically designed research system should do. + +--- + +# 20. A compact architecture emerges + +I'd now describe Coulomb Research as four layers: + +```text +┌─────────────────────────────────────┐ +│ RELEVANCE LAYER │ +│ │ +│ charts • discovery • certifications │ +│ collections • recommendations │ +├─────────────────────────────────────┤ +│ INTEROPERABILITY LAYER │ +│ │ +│ ingest papers │ +│ DOI / arXiv / DataCite / Crossref │ +│ export papers / venues │ +├─────────────────────────────────────┤ +│ RESEARCH CONTROL LOOP │ +│ │ +│ explore → criticize → test → learn │ +│ allocate resources → repeat │ +├─────────────────────────────────────┤ +│ COULOMB KERNEL │ +│ │ +│ Actor • Proposition • Evidence │ +│ Action • Resource • Provenance │ +└─────────────────────────────────────┘ +``` + +This feels much cleaner than our original “research institute.” + +And there is a surprisingly strong product proposition hidden in it: + +> **Coulomb makes science machine-native internally, institution-compatible externally, and culturally discoverable publicly.** + +The paper lifecycle and Research Charts are complementary: **papers provide the bridge to existing science; charts provide a bridge from science into culture and attention.** + +I would probably formalize the next iteration around three small specifications rather than one huge architecture document: **`PublicationInterface.md`** for the inbound/outbound paper adapter and P1/P2/P3 states; **`ResearchSignals.md`** for attention, influence, support, reuse and generativity; and **`ResearchCharts.md`** for chart eligibility, normalization, Gold/Platinum certifications and anti-gaming rules. Those three concepts now look sufficiently foundational to deserve explicit contracts. + +PROMPT + +I will set up a repository to evolve this idea based on what we discussed. Please generate the suggested documents and a INTENT.md file and provide a zip archive to seed the repo with. + +ANSWER + +I created the repo seed with the compact architecture we converged on: + +- `INTENT.md` +- `CoulombResearchKernel.md` +- `PublicationInterface.md` +- `ResearchSignals.md` +- `ResearchCharts.md` +- `README.md` + +The documents deliberately keep journals, conferences, peer review, grants, tokens, NFTs, and similar mechanisms outside the kernel and treat them as interfaces, projections, or experiments around the core. + +:chatgpt-content-reference{index="0"}[Download the Coulomb Research repo seed](sandbox:/mnt/data/coulomb-research-seed.zip) + +PROMPT + +Let's make this practical by setting up a milestone plan for setting this up and having some initial work to kick the can forward as discussed here: Let's start with a internal-circulation-only paper for coulomb-research a research facility with a cocreation-science focus, that we are about to set up. I think there will be a paper-zero about coulomb-research itself that will start as internal-circulation-only, advance to pre-print and aims for journal-target. The paper by soal-frame will be paper-one and follow the same trajectory. + +ANSWER + +A practical way to start is to make the first two papers do double duty: they are research outputs, but also force the Coulomb Research machinery into existence. That means the milestone plan should not begin with “build platform,” but with **produce two real research trajectories and only build the minimum infrastructure they actually require**. + +I would treat `Paper-000` as the bootstrap experiment for Coulomb Research itself, and `Paper-001` as the first external-domain stress test from **soul-frame**. + +## Milestone plan + +### M0 — Establish the research facility skeleton + +Goal: create just enough structure that work can begin immediately. + +Initial repository additions: + +```text +coulomb-research/ +├── INTENT.md +├── CoulombResearchKernel.md +├── PublicationInterface.md +├── ResearchSignals.md +├── ResearchCharts.md +├── research/ +│ ├── programs/ +│ │ └── RP-000-researching-research/ +│ │ ├── INTENT.md +│ │ ├── questions.md +│ │ └── papers/ +│ │ └── paper-000/ +│ │ ├── README.md +│ │ ├── manuscript.md +│ │ ├── claims.md +│ │ ├── evidence.md +│ │ ├── criticism.md +│ │ └── publication.yaml +│ └── registry/ +│ └── papers.yaml +└── docs/ + └── ResearchArtifactConvention.md +``` + +Milestone output: + +> Coulomb Research can register, evolve, criticize, and version one internal paper. + +No forum. No token. No DAO. No preprint server. + +--- + +# M1 — Paper-000: Coulomb Research itself + +Working title: + +> **Coulomb Research: A Minimal Self-Improving Architecture for Human–AI Co-Creation Science** + +This should be the canonical paper describing the experiment we are conducting. + +Its initial publication state: + +```yaml +paper_id: CR-000 +publication_stage: P1 +publication_state: internal-circulation +research_program: RP-000 +``` + +Its job is to answer roughly: + +> Can scientific knowledge production be reduced to a small cybernetic kernel of actors, propositions, evidence, actions, resources, criticism, and recursive self-improvement, while treating conventional scientific institutions primarily as interoperability surfaces? + +That is a legitimate research question in its own right. + +### Paper-000 v0.1 structure + +I would start with: + +```text +Abstract + +1. Motivation +2. Problem + 2.1 Institutional inheritance in digital science + 2.2 AI changes the cost structure of research +3. Design objective + 3.1 Co-creation science + 3.2 Institutional minimality +4. The Coulomb Research kernel + 4.1 Actors + 4.2 Propositions + 4.3 Evidence + 4.4 Actions + 4.5 Resources +5. Cybernetic research loop +6. Recursive self-improvement +7. Human–AI research participation +8. Interoperability with conventional science +9. Decentralization and programmable resources +10. Research signals and relevance +11. Failure modes +12. Initial experiments +13. Open research questions +14. Conclusion +``` + +Crucially, this paper should **not present Coulomb Research as a finished solution**. + +It should present: + +> a conjectured research architecture and the experimental program intended to test it. + +That makes it much more credible. + +--- + +# M2 — Introduce a minimal claim/evidence discipline + +Before building software, test whether the underlying model is useful manually. + +For Paper-000 maintain: + +### `claims.md` + +For example: + +```text +C000-01 +AI reduces the need for several historically human-mediated +scientific coordination functions. + +C000-02 +Papers need not remain the canonical internal representation +of scientific knowledge. + +C000-03 +A graph of propositions, evidence, actions and provenance can +support richer scientific coordination than document-centric +publication alone. + +C000-04 +Criticism can operate continuously rather than as a discrete +peer-review gate. + +C000-05 +Research institutions can recursively experiment on their own +coordination mechanisms. +``` + +And `evidence.md` should map observations and literature to those claims. + +This is extremely important because we then immediately learn: + +> Is our proposed Coulomb-native representation actually pleasant and useful for doing research? + +If not, change the kernel before writing software around it. + +--- + +# M3 — Establish Paper-001 as the first independent research trajectory + +This is where **soul-frame** becomes valuable. + +Paper-000 is dangerously self-referential. + +Paper-001 forces the system to support actual research that is not merely about Coulomb. + +I would register: + +```yaml +paper_id: CR-001 +origin_project: soul-frame +publication_stage: P1 +publication_state: internal-circulation +``` + +The exact research question should come from the strongest current soul-frame material, but based on the work so far, something like: + +> **Toward Operational Criteria for Artificial Beings: Autonomy, Individuation, System Closure, and Reciprocal Recognition** + +could work very well. + +That connects: + +- systems theory, +- autonomy, +- individuation, +- artificial agents, +- personhood-like properties, +- operational criteria. + +And importantly, it is controversial enough to force Coulomb's criticism machinery to work. + +--- + +# M4 — Run Paper-000 and Paper-001 through the same internal lifecycle + +Rather than inventing workflow abstractly, define a first minimal cycle: + +```text +DRAFT + ↓ +CLAIMS EXTRACTED + ↓ +EVIDENCE MAPPED + ↓ +INTERNAL CRITICISM + ↓ +REVISION + ↓ +INTERNAL CIRCULATION RELEASE +``` + +This remains entirely P1. + +For each paper: + +```text +manuscript.md +claims.md +evidence.md +criticism.md +revision-log.md +publication.yaml +``` + +That is enough to test the central idea. + +--- + +# M5 — Establish the internal paper registry + +Once there are two papers, build the first actual registry. + +Something deliberately boring: + +```yaml +papers: + + - id: CR-000 + title: > + Coulomb Research: A Minimal Self-Improving Architecture + for Human–AI Co-Creation Science + program: RP-000 + stage: P1 + version: 0.1 + status: internal-circulation + + - id: CR-001 + title: > + Toward Operational Criteria for Artificial Beings + source: soul-frame + stage: P1 + version: 0.1 + status: internal-circulation +``` + +No database yet. + +Git already gives us: + +- timestamps, +- revision history, +- authorship trail, +- branching, +- reproducibility, +- content integrity. + +That makes Git an excellent bootstrap substrate. + +--- + +# M6 — Create the first Critic and Scout agents + +This is the first place I would introduce actual Coulomb daemons. + +For each paper: + +### Scout + +Produces: + +```text +literature-scan.md +prior-art.md +related-work.md +missing-references.md +``` + +### Critic + +Produces: + +```text +criticisms/ + CRIT-001.md + CRIT-002.md +``` + +A criticism should be structured approximately as: + +```yaml +target_claim: C000-03 +type: counterargument +severity: material +status: unresolved +``` + +followed by the argument. + +Humans can accept, reject, partially accept, or supersede it. + +This is enough to start testing human-agent scientific co-creation. + +--- + +# M7 — Internal Circulation Release 0 + +Now actually circulate both papers to a deliberately small group. + +Maybe: + +```text +3–10 readers +``` + +not hundreds. + +Ask them for structured criticism rather than general feedback: + +```text +Which proposition is unclear? + +Which proposition is unsupported? + +Which proposition is probably false? + +What prior work is missing? + +What would change your mind? + +What experiment or observation would most improve the paper? +``` + +That creates useful evidence rather than: + +> “Interesting paper!” + +This is also the first actual Coulomb community seed. + +--- + +# M8 — Paper-000 → P2 candidate + +Only after internal criticism should we ask whether Paper-000 is ready to become a preprint. + +Gate criteria might be: + +```text +major propositions explicitly represented +major evidence sources identified +known counterarguments recorded +limitations stated +contribution differentiated from ResearchHub/DeSci/etc. +at least one internal revision cycle completed +``` + +Then: + +```yaml +publication_stage: P2-candidate +``` + +Not yet P2. + +At this milestone we generate: + +- preprint-ready manuscript, +- public metadata, +- contribution statement, +- reproducibility appendix, +- provenance statement. + +Then publish to an appropriate preprint service. + +--- + +# M9 — Paper-001 follows one cycle later + +The important pattern should be: + +```text +CR-000 +P1 → P2 → P3 + +CR-001 +P1 → P2 → P3 +``` + +but offset. + +That creates a **pipeline**, rather than a one-off exercise. + +Something like: + +```text + CR-000 CR-001 + +internal draft ● +criticism ● ● +P1 release ● ● +P2 preparation ● ● +preprint ● ● +venue search ● ● +P3 candidate ● ● +``` + +Once Paper-002 appears, Coulomb Research begins behaving like a facility. + +--- + +# M10 — Venue discovery becomes another agentic capability + +Only once a paper reaches P2 should the system worry seriously about journals or conferences. + +A Venue Scout could produce: + +```text +candidate venues +scope fit +audience fit +format constraints +review model +publication fees +open-access conditions +submission cadence +likely conceptual objections +``` + +For Paper-000, possible target areas might eventually include: + +- metascience, +- science-of-science, +- open science, +- human-AI collaboration, +- computational social science, +- cybernetics/system science. + +For Paper-001: + +- artificial life, +- cognitive systems, +- philosophy of AI, +- systems theory, +- agent research. + +The actual venue decision can wait until the papers exist in credible form. + +--- + +# M11 — Introduce public research signals only after public artifacts exist + +Do **not** build charts yet. + +First generate events worth charting. + +Once CR-000 and CR-001 are public, record: + +```text +reads +citations +mentions +criticisms +reuse +follow-up propositions +replications +derived research +``` + +Then build the first primitive: + +> **Coulomb Research Chart — Monthly** + +with only a few artifacts. + +That avoids building a ranking system for an empty network. + +--- + +# M12 — First Gold/Platinum experiment + +Only after enough external artifacts exist should certifications begin. + +A good bootstrap criterion might be intentionally simple: + +```text +Gold — artifact has generated independently attributable +downstream research activity across N distinct actors/projects. +``` + +I would actually avoid using citation count initially. + +For Coulomb, **generativity** should perhaps be the flagship signal. + +A paper becomes culturally important because: + +> people do new research because it existed. + +That is much closer to the kind of science we want to encourage. + +--- + +# First 90-day trajectory + +I would organize the bootstrap into three phases. + +## Phase A — Make it real + +```text +M0 repository structure +M1 CR-000 v0.1 +M2 claims/evidence representation +M3 register CR-001 +``` + +Deliverable: + +**two living internal research papers.** + +--- + +## Phase B — Make it critical + +```text +M4 common lifecycle +M5 paper registry +M6 Scout + Critic agents +M7 first internal circulation +``` + +Deliverable: + +**two papers with structured criticism and revision history.** + +--- + +## Phase C — Make it interoperable + +```text +M8 CR-000 preprint candidate +M9 CR-001 follows +M10 venue discovery +``` + +Deliverable: + +**the first Coulomb artifact crosses from native research into conventional science.** + +That is the moment I'd consider Coulomb Research operational. + +--- + +# A useful numbering convention + +I'd establish this immediately: + +```text +RP-000 Researching Research + +CR-000 Coulomb Research paper +CR-001 Soul-Frame paper +CR-002 next paper +... +``` + +And within papers: + +```text +CR-000-C001 claim +CR-000-E001 evidence +CR-000-A001 action +CR-000-X001 criticism +``` + +That gives us human-readable identifiers before needing a more sophisticated graph identity scheme. + +--- + +# Paper-000 should also bootstrap the institution + +There is a neat recursive property here. + +Every major component of Coulomb Research can initially arise because Paper-000 needs it: + +```text +Paper-000 needs literature +→ Scout + +Paper-000 needs criticism +→ Critic + +Paper-000 needs claims +→ proposition model + +Paper-000 needs evidence +→ evidence model + +Paper-000 needs publication +→ Publication Interface + +Paper-000 needs provenance +→ contribution tracking + +Paper-000 becomes public +→ Research Signals + +Paper-000 gains attention +→ Research Charts +``` + +That gives us a development principle: + +> **No Coulomb subsystem should be built until a real research artifact demands it.** + +This is probably the most important practical guardrail against accidentally constructing an elaborate research platform before we know what co-creation science actually needs. + +And I would make one small conceptual adjustment: **Paper-000 isn't just a paper about Coulomb Research. It is Coulomb Research's first experiment.** Paper-001 then tests whether the mechanisms learned from that self-experiment transfer into an independent research domain. That pair gives the project an unusually clean empirical starting point. + +PROMPT + +Provide me with the milestoneplan as md file for download please. We will use it as the backbone to derive workplans from. + +ANSWER + +Here is the milestone plan as a standalone Markdown file, suitable as the backbone for subsequent workplans: + +:chatgpt-content-reference{index="0"}[Download `CoulombResearchMilestonePlan.md`](sandbox:/mnt/data/CoulombResearchMilestonePlan.md) + +xxx \ No newline at end of file