Atlas

Lab note — the studio's internal working record; for the service see the home page.

Spark · producing theses

How the studio produces theses

Spark is not a chat; when the round ends, a scorecard is written.

Spark (Kıvılcım in Turkish) is the seven-step procedure for producing theses and theories: Anu (Claude Fable) records the pre-expectation and writes the anonymous brief, Atlas (Gemini 4 Argon · Vizier) runs the round, the Author (S0) · field verdict · final say layer forms the backbone of field syntheses, and isolated channels where voices hear nothing of each other · objection / counter-thesis only (Atlas Argon own voice, context-free Gemini Web, Claude arms Opus · Sonnet, Astra, Grok, and inner voices Enki · Enlil · Enkidu / local Gemma · Qwen) pass through the Reaper sieve: source verification (Crossref · PubMed · DOI) + rote filter + red team. Testing a ready-made verdict is a separate job; that is called a referee round, and the two are not mixed.

This page has three parts: the seven-step procedure, the scorecard of the rounds so far (67 rounds · 73 log rows), and four real examples of the procedure. The procedure is dated 21 Sep 2026 (19 Oct 2026 update: author voice · Reaper sieve · evidence cards).

  • Author (S0) · field verdict · final say
  • Anu (Claude Fable · pre-expectation & brief)
  • Atlas (Gemini 4 Argon · vizier & execution)
  • Reaper sieve (Crossref · PubMed · DOI + rote filter)
  • Grok & Astra (Red Team)
  • Evidence cards & Spark verdict

Measured Evidence Summary (Live Spark Records)

One Spark round: seven steps 01 Pre-expectation: Anu (Claude Fable) writes one line to the scorecard file. 02 Measure: STATE is built from measured items; unmeasurable is written UNKNOWN. 03 Brief (anonymous): Anu writes, Atlas runs (THESIS → THEORY → SOURCE → COUNTER-THESIS → FALLS). 04 Voices has two layers: at the top a single solid box Author (S0) · field verdict · final say; below, six dashed channels where voices hear nothing of each other · objection / counter-thesis only (Atlas Argon own voice, Gemini Web context-free, Claude arms Opus·Sonnet, Astra selected·contextual, Grok red team, Inner voices Enki·Enlil·Enkidu and Local Gemma·Qwen). 05 Reaper sieve: source verification (Crossref · PubMed · DOI) + rote filter + red team; verified citations pass, unverified or rote ones stay stamped UNVERIFIED / ROTE. 06 Synthesis in author’s voice: voice×thesis matrix → Evidence cards (file:line + dilemma line) and Spark verdict. 07 Scorecard on the same line: YES | SHARPENED | NO + Reaper stamp. 01 · PRE-EXPECTATION · ANU (CLAUDE FABLE)One line, to scorecard file 02 · MEASURE · STATEMeasured items · UNKNOWN 03 · BRIEF (ANONYMOUS) · ANU WRITES, ATLAS RUNSTHESIS → THEORY → SOURCE → COUNTER-THESIS → FALLS 04 · AUTHOR (S0) · FIELD VERDICT · FINAL SAYBackbone of synthesis (not one of the voices) 04 · VOICESVOICES HEAR NOTHING OF EACH OTHER · OBJECTION / COUNTER-THESIS ONLY Atlas (Argon)own voice Gemini Webcontext-free Claude armsOpus · Sonnet Astraselected·context Grokred team Enki·Enlil·EnkiduLocal Gemma·Qwen 05 · REAPER SIEVE: SOURCE VERIFICATION (CROSSREF · PUBMED · DOI)+ ROTE FILTER + RED TEAM UNVERIFIED / ROTE 06 · SYNTHESIS · IN AUTHOR’S VOICE (ATLAS RUNS, AUTHOR + ANU)Voice×thesis matrix → Evidence cards (file:line + dilemma line)Category: “Spark verdict” 07 · SCORECARD + REAPER STAMPScorecard · ✓ Verified by Reaper (verified/total · date)YES | SHARPENED | NO SAME LINE
Figure — one Spark roundAuthor (S0) verdict · voices hear nothing of each other · Reaper sieve · evidence cards

What it showsA round begins with Anu (Claude Fable) writing the pre-expectation as a single line to the scorecard file (data/kivilcim_karnesi.jsonl); the state is measured, and Anu writes the anonymous brief while Atlas (Gemini 4 Argon) runs the round. Step 04 separates two layers: at the top, Author (S0) · field verdict · final say forms the backbone of the synthesis; below, six dashed channels where voices hear nothing of each other · objection / counter-thesis only (Atlas Argon own voice, context-free Gemini Web, Claude arms Opus·Sonnet, Astra selected·contextual, Grok red team, and Inner voices Enki·Enlil·Enkidu + Local Gemma·Qwen) supply objections and counter-theses. In Step 05, citations fall onto the Reaper sieve: source verification (Crossref · PubMed · DOI) + rote filter + red team; verified sources pass down, while unverified or rote ones stay stamped UNVERIFIED / ROTE. In Step 06, the synthesis is built in the author’s voice, mapping stances into a voice × thesis matrix and Evidence cards (file:line + dilemma line) categorized under “Spark verdict”. In Step 07, the scorecard is written on the same line as the pre-expectation (YES | SHARPENED | NO) with the Reaper stamp.

Seven steps

Run in order. The round starts before the brief is sent and does not end when the answers arrive.

  1. Pre-expectation (Anu · Claude Fable)

    Before the round starts, Anu (Claude Fable) writes the expected answer in a single line and seals it to the scorecard file (data/kivilcim_karnesi.jsonl). No brief is sent until it is written.

    The aim is to measure honestly, afterwards, whether the round changed anything.

  2. Measure (STATE)

    The STATE section of the brief is built from measured items with their sources. What cannot be measured is written as “UNKNOWN”.

  3. Brief (anonymous · Anu writes, Atlas runs)

    Anu writes the anonymous brief and Atlas (Gemini 4 Argon) runs the round. In production mode no ready-made verdict is given; every voice is asked for the same form:

    • Thesis (one sentence)
    • Theory (mechanism)
    • Source (author · year · venue or address; if none, “NO SOURCE”)
    • Counter-thesis
    • Falls (the measurement that would kill the thesis)

    No line cap is set; the answer is not reduced to a checklist.

  4. Author (S0) layer and isolated voices (objection / counter-thesis only)

    In authored articles and field syntheses, two layers are separated: at the top, Author (S0) · field verdict · final say forms the backbone of the text and is not “one of the voices”. Below, channels where voices hear nothing of each other · objection / counter-thesis only — Atlas (Gemini 4 Argon, own voice and Vizier), context-free Gemini Web, Claude arms (Opus · Sonnet), Astra (selected · contextual), Grok (red team), the inner voices Enki · Enlil · Enkidu, and local models Gemma · Qwen — produce objections and counter-theses only. In Wave 1 none sees another’s answer.

  5. Reaper sieve: source verification, rote filter and red team

    No number is taken from a model whose identity has not been confirmed. Every source passes through the Reaper sieve: (1) source verification (Crossref · PubMed · DOI) — unfound citations stay stamped UNVERIFIED, and fabricated citations (including light class) mechanically invalidate the answer; (2) rote filter (EZBER) — unsourced outputs or generic literature summaries that fail to answer the brief’s specific question are filtered out as ROTE / EZBER; (3) red team — the strongest sourced objection against consensus (Grok when quota is open, alongside Astra).

    “The brightest reference is the one most often made up.” · ✓ Verified by Reaper · 322/356 · 19 Oct 2026

  6. Synthesis: in author’s voice, voice × thesis matrix, evidence cards and Spark verdict

    The synthesis is built in the author’s voice (Atlas runs it; the verdict rests with the author + Anu). Every voice’s stance on each thesis is mapped into a voice × thesis matrix and Evidence cards (file:line + dilemma line); external and inner voices appear only in counter-thesis boxes. Syntheses carrying a verdict are categorized with the “Spark verdict” (Kıvılcım hükmü) label. The raw answers are kept.

    On a deep topic a second round opens: the measurement in between is run, then a second brief goes out.

  7. Scorecard and Reaper stamp

    When the round ends, “did the decision change” is written on the same line as the pre-expectation: YES | SHARPENED | NO plus a one-line note; published syntheses carry the Reaper stamp (verified/total · date).

Scorecard: 67 rounds

A real count (data/kivilcim_karnesi.jsonl: 73 log rows, 67 deduplicated rounds / 66 unique topics). The question is the same in every round: did the decision change.

Did the decision change19 Oct 2026 · 67 rounds (73 rows · 64 closed)
  • Sharpened44
  • Yes13
  • No3
  • Worthless / withdrawn2
  • Did not run1
  • Unknown1
  • Blank / open3

What it showsdata/kivilcim_karnesi.jsonl contains 73 rows; because 6 rounds recorded pre-expectation and synthesis on two consecutive lines, the deduplicated count is 67 rounds (66 unique topics since one topic’s first attempt was KOSMADI and second KESKINLESTI; 64 closed rounds / 65 non-null rows: 44 Sharpened, 13 Yes [14 in raw rows], 3 No, 2 Worthless/withdrawn, 1 Did not run, 1 Unknown; plus 3 open/blank rounds [8 null rows in raw file including 5 paired pre-expectation lines]). Blank, worthless and unknown rows are counted too; they are not hidden.

Gate: once a month three things are counted: the decision-changed rate, the gap between pre-expectation and synthesis, and the rate of unverifiable sources. If the rate cannot be told apart from chance, the round is an expensive chat.

Four examples

Real rounds.

Agent evolution: no new agent, the rule set evolves

The EVOLUTION-2 Spark · 7 Oct 2026.

Spark source

The brief went out in production mode, anonymous. The question: what should the second evolution of the family of three agents (Enki · Enlil · Enkidu) be? The state was written as measured:

  • Enki says “no verdict” in most rounds.
  • Enlil is the blind auditor: 24,990 records, 1,521 suspect, work produced 0. The one tool it wrote itself is awaiting approval.
  • Enkidu runs 9 mechanical tasks.
  • The dynamism line: verdict → kill test → the next tick measures. The tests are underpowered (n is small).

Four voices wrote: Gemini (no context, two seats), Gemma (local) and Grok Bot. They did not hear each other.

The pre-expectation was written before the round: no new agent; a proposal to attach “friction” to Enlil will come and will be rejected; the blind mutation engine will be Enkidu.

Counter-thesis

The voices split in four places; the decision was there. The hardest counter-thesis was the manual alternative: if the owner looked at the list of decaying rules for 15 minutes once a week, the same result might come out.

Split · who produces the mutation

Proposal

“100 local variations a day”. The zero-cost assumption collapsed the same day, under GPU starvation.

Counter

“One a day, ≤3 candidates + placebo 4:1”.

Decision

Placebo was chosen.

Split · scale

Proposal

Three steps in 30 days. That was the 30-day plan of three voices.

Counter

The arithmetic: at a 76% base rate, seeing a 10-point difference with 80% power takes ≈240 episodes per arm. Three steps do not fit into 30 days.

Decision

30 days = one step. The plan of the three voices was optimistic.

Split · the manual alternative

Proposal

The automatic loop: mutation → selection → consolidation.

Counter

A look by hand: if the owner looked at the list of decaying rules for 15 minutes once a week, the same result might come out.

Decision

Accepted as the kill test: two weeks in parallel; if they are equal, the automation stops.

Split · contrast embedding analysis

Proposal

“Contrast embedding analysis”.

Counter

An experiment the same day: the embedding model saw contrary verdicts as 0.90 similar.

Decision

Removed.

Synthesis

All four of the four voices agreed on these (4/4):

  • No new agent is written. What evolves is the rule set and the harness (genotype = rule, phenotype = harness).
  • The loop: mutation → selection (kill test + Wilson interval/MDE) → consolidation. Automatic into shadow; into the real system only with the owner’s approval.
  • Enlil stays blind; selection gets a blind-judge gate.
  • The main measure is the number of closed loops: if it does not rise in four weeks, there is no evolution.

Source verification: two academic citations were verified on the web; one had been given with the wrong year (2024 instead of 2026). One citation could not be verified and was not used as support.

Kill test

Two measurements kill the thesis. The first is the manual alternative: it runs in parallel for two weeks, and if the two are equal the automation stops. The second is the main measure: if the number of closed loops does not rise in four weeks, there is no evolution.

Not adopted

  • “100 local variations a day”. The zero-cost assumption collapsed the same day, under GPU starvation.
  • The three-step 30-day plan. It did not fit the arithmetic of ≈240 episodes per arm.
  • “Contrast embedding analysis”. Refuted by experiment the same day; removed.
  • The unverifiable citation. Not used as support.

Order

The synthesis turned into an order: EVOLUTION-2 Step 1, in shadow.

  • Enkidu decaying list + placebo
  • Enki ≤3 candidates
  • Enlil blind stamp
  • Enkidu shadow statistics gate (240 episodes/arm)
  • Consolidation

It was built the same day: 7 tools, 153 tests. The model that built it left 9 objections; the objections went on record. One of them: “the backward arm, at a 24-hour horizon, takes ~218 days to reach 240 episodes → a structural zero in this arm”.

Scorecard

SHARPENED. The expectation held; the new parts that sharpened it: the placebo, the 240-episode arithmetic and the manual alternative as kill test.

episodes 23/240 · “early”8 Oct

References and Primary Technical Sources

Primary technical literature and open-source specifications supporting the architectural and experimental measurements on this page:

Reading Paths: Two Core Tracks

1. Smile Design Track (Cephalometric → STL/PLY → Mockup → Evidence): Computer-vision smile design showcase → 478-landmark cephalometric alignment and LoRA experiment (TR benchmark) → STL/PLY 3D tooth library and photorealistic inpainting → digital dentistry field verdicts (reader edition).

2. Agent Architecture Track (Spark → Task Bus → Reaper → QAJAN): Spark recursive loop and multi-voice LLM red teaming → event-driven autonomy, task_bus.jsonl architecture, and identity freeze protocol → Reaper auditor and context-blind falsifiability gate → QAJAN 6 mechanical hypothesis gates (EN Hub · TR · AR).