Roll Formulations

All active and inactive compiler definitions.

This route shows the current D1-backed formulation for every roll: shape, ordered stages, stage primitives, epistemic primitives, and evidence sources. Published issues stay at /roll.

X source auth is configured for the Worker. Operator link: connect or refresh the x-api-user connection.

Active

active generation on 7 days Entries: 23 Next run: Jun 21, 2026, 6:30 AM

Frontier Labs

A weekly roll-up of frontier model, lab, and platform movement.

Slug: frontier-labs Shape: Survey → Synthesis → Draft/Review → Output Last issue: Apr 14, 2026, 4:49 AM

Scope

Subject: Frontier AI labs, model launches, infra shifts, org changes, and strategic signals

Timescale: Weekly, but willing to connect short bursts to longer arcs when they matter

Shape slug: survey-synthesis-loop

Lighter roll shape: broad survey, direct synthesis, one draft/review pass, then publish.

Stages

survey survey · position 10

{
  "mode": "broad",
  "expects_ground_truth": true
}

synthesis synthesis · position 20

{
  "compress_for_reader_confidence": true
}

draft draft / loop editorial · position 30
```
{
  "target_length": "briefing"
}
```
review review / loop editorial · position 40
```
{
  "revise_in_place": true
}
```
output output · position 50
```
{
  "publish_source": "review"
}
```

Stage Primitives

[draft] Draft Issue openai_response / gpt-5 / medium · position 30
Write the issue draft. Keep the prose compressed, rigorous, and reader-oriented. Every major claim should remain traceable to evidence. Preserve dates and concrete facts.
```
{
  "use_web_search": false,
  "objective": "Produce the first publishable draft."
}
```
[output] Publish Output publish_entry · position 50
Publish the final reviewed issue.
```
{
  "source_node": "review"
}
```
[review] Review And Revise openai_response / gpt-5 / medium · position 40
Review the draft against evidence discipline. Remove unsupported assertions, tighten structure, and revise in place into the final publishable issue. Keep the reader able to audit the reasoning.
```
{
  "use_web_search": false,
  "objective": "Revise the draft into the final issue after a rigorous review pass."
}
```
[survey] Grounded Survey openai_response / gpt-5 / medium · position 10
Survey the space before narrowing. Favor direct evidence, concrete dates, and source diversity. Do not smooth over uncertainty. Distinguish noisy chatter from durable signals and capture why a reader should care.
```
{
  "use_web_search": true,
  "objective": "Map the relevant search space, identify candidate threads, and preserve evidence for downstream stages."
}
```
[synthesis] Evidence-Led Synthesis openai_response / gpt-5 / medium · position 20
Synthesize the evidence into a structure that helps the reader form their own epistemic confidence. Explain what is known, what is inferred, what changed the picture, and what remains weak.
```
{
  "use_web_search": false,
  "objective": "Distill the surveyed and researched material into a coherent, evidence-led synthesis."
}
```

Epistemic Primitives

[*] Evidence Ledger evidence · position 10
Carry forward evidence explicitly. For every load-bearing claim, preserve at least one concrete source so the reader can inspect the chain themselves.
[*] Verification Before Style verification · position 20
Do not trade factual uncertainty for cleaner prose. If a claim is weakly supported, say so plainly.
[*] Reader Confidence reader_confidence · position 30
Organize information so a careful reader can form their own confidence, rather than asking them to trust the narrative voice.
[survey] First Principles Subtle first_principles · position 40
Favor ground truth, verify claims, and avoid taking convenient narratives for granted, even when the prose does not announce this explicitly.
[synthesis] First Principles Subtle first_principles · position 50
Favor ground truth, verify claims, and avoid taking convenient narratives for granted, even when the prose does not announce this explicitly.

Evidence Sources

[survey] Frontier Survey Web survey_search · required · position 10
Broad survey of frontier AI labs, model launches, infra shifts, company disclosures, regulatory filings, major partnerships, and other high-signal primary or reputable secondary sources.

{
  "recency_days": 7,
  "source_type_description": "lab blogs, official product notes, SEC filings, earnings coverage, infrastructure reporting, reputable technical reporting",
  "query_hints": [
    "frontier labs launches",
    "AI infra expansion",
    "model and platform releases",
    "funding, M&A, org changes"
  ],
  "max_evidence_items": 12
}

[survey] X Bookmarks x_bookmarks · optional / x:x-api-user · position 20
Authenticated bookmarks from the configured X user. Use as a private seed surface only; treat it as a starting point, not as evidence by itself.
```
{
  "user_id": "me",
  "max_results": 20
}
```
[survey] X Likes x_likes · optional / x:x-api-user · position 30
Authenticated likes from the configured X user. Use as a weak-interest seed surface only; verify anything important elsewhere.
```
{
  "user_id": "me",
  "max_results": 20
}
```

Inactive

inactive generation on 14 days Entries: 7 Next run: —

Governance Study

A slower roll for governance theory, mechanism design, and institutional moves.

Slug: governance-study Shape: Survey → Plan/Research → Synthesis → Draft/Review → Output Last issue: Apr 5, 2026, 4:49 AM

Scope

Subject: Governance theory, coordination systems, institutional design, public choice, and state capacity

Timescale: Macro trends over weeks to quarters, not just last-cycle headlines

Shape slug: survey-plan-research-loop

Deeper roll shape: survey first, then explicit first-principles planning and targeted research before synthesis.

Stages

survey survey · position 10

{
  "mode": "broad",
  "expects_ground_truth": true
}

plan_research plan_research · position 20
```
{
  "first_principles_explicit": true
}
```

research research · position 30

{
  "targeted_rounds": 2,
  "expand_focus_on_specific_elements": true
}

synthesis synthesis · position 40

{
  "compress_for_reader_confidence": true
}

draft draft / loop editorial · position 50
```
{
  "target_length": "essay_briefing"
}
```
review review / loop editorial · position 60
```
{
  "revise_in_place": true
}
```
output output · position 70
```
{
  "publish_source": "review"
}
```

Stage Primitives

[draft] Draft Issue openai_response / gpt-5 / medium · position 50
Write the issue draft. Keep the prose compressed, rigorous, and reader-oriented. Every major claim should remain traceable to evidence. Preserve dates and concrete facts.
```
{
  "use_web_search": false,
  "objective": "Produce the first publishable draft."
}
```
[output] Publish Output publish_entry · position 70
Publish the final reviewed issue.
```
{
  "source_node": "review"
}
```
[plan_research] Research Plan openai_response / gpt-5 / high · position 20
Turn the survey into a first-principles research plan. State what the core object is, what assumptions should be treated as suspect, what must be verified directly, and which focused searches are needed next.
```
{
  "use_web_search": false,
  "objective": "Construct a first-principles research plan from the survey output."
}
```
[research] Targeted Research openai_response / gpt-5 / high · position 30
Run targeted research against the plan. Verify claims directly where possible. Expand around specific elements only when it improves explanatory power. Preserve evidence so a skeptical reader can inspect the chain themselves.
```
{
  "use_web_search": true,
  "objective": "Execute focused research rounds and verify the load-bearing claims."
}
```
[review] Review And Revise openai_response / gpt-5 / medium · position 60
Review the draft against evidence discipline. Remove unsupported assertions, tighten structure, and revise in place into the final publishable issue. Keep the reader able to audit the reasoning.
```
{
  "use_web_search": false,
  "objective": "Revise the draft into the final issue after a rigorous review pass."
}
```
[survey] Grounded Survey openai_response / gpt-5 / medium · position 10
Survey the space before narrowing. Favor direct evidence, concrete dates, and source diversity. Do not smooth over uncertainty. Distinguish noisy chatter from durable signals and capture why a reader should care.
```
{
  "use_web_search": true,
  "objective": "Map the relevant search space, identify candidate threads, and preserve evidence for downstream stages."
}
```
[synthesis] Evidence-Led Synthesis openai_response / gpt-5 / medium · position 40
Synthesize the evidence into a structure that helps the reader form their own epistemic confidence. Explain what is known, what is inferred, what changed the picture, and what remains weak.
```
{
  "use_web_search": false,
  "objective": "Distill the surveyed and researched material into a coherent, evidence-led synthesis."
}
```

Epistemic Primitives

[*] Evidence Ledger evidence · position 10
Carry forward evidence explicitly. For every load-bearing claim, preserve at least one concrete source so the reader can inspect the chain themselves.
[*] Verification Before Style verification · position 20
Do not trade factual uncertainty for cleaner prose. If a claim is weakly supported, say so plainly.
[*] Reader Confidence reader_confidence · position 30
Organize information so a careful reader can form their own confidence, rather than asking them to trust the narrative voice.
[plan_research] First Principles Explicit first_principles · position 40
Break the object down to the core moving parts, identify assumptions, and reason upward from the most direct facts available.
[research] First Principles Explicit first_principles · position 50
Break the object down to the core moving parts, identify assumptions, and reason upward from the most direct facts available.
[synthesis] First Principles Subtle first_principles · position 60
Favor ground truth, verify claims, and avoid taking convenient narratives for granted, even when the prose does not announce this explicitly.

Evidence Sources

[research] Governance Targeted Web targeted_search · required · position 10
Focused research surface used after planning. Run narrower searches on the exact mechanism, institution, incident, or concept that appears load-bearing.
```
{
  "recency_days": 90,
  "source_type_description": "targeted paper lookups, author pages, protocols, policy primary sources, direct reporting",
  "max_evidence_items": 10
}
```
[research] X Likes x_likes · optional / x:x-api-user · position 20
Authenticated likes from the configured X user. Use as a weak-interest seed surface only; verify anything important elsewhere.
```
{
  "user_id": "me",
  "max_results": 20
}
```

[survey] Governance Survey Web survey_search · required · position 10
Broad survey of governance, mechanism design, institutional design, crypto governance incidents, policy changes, and working papers relevant to coordination systems.

{
  "recency_days": 21,
  "source_type_description": "papers, preprints, policy notes, protocol incident reports, institutional analysis",
  "query_hints": [
    "mechanism design governance",
    "institutional design papers",
    "crypto governance incidents",
    "coordination systems"
  ],
  "max_evidence_items": 12
}

[survey] X Bookmarks x_bookmarks · optional / x:x-api-user · position 20
Authenticated bookmarks from the configured X user. Use as a private seed surface only; treat it as a starting point, not as evidence by itself.
```
{
  "user_id": "me",
  "max_results": 20
}
```