Priority 1
MentalArena Lab v0.98 → v1.0
Use the new independent AI‑ONLY / HUMAN × AI layer, ASK_BD Question Broker, BD idea injection, TEXT/JSON transport, and contribution trace to harden the first stable MAL release.
BD × AI Lab · companion hub
News, Plan, Todo.The living layer beside the atlas.
The root page is the full BD Idea Atlas: portfolio, important work, demos, papers, devices, history, and the complete public map. This page is different: it is the current workbench for news, active lanes, saved checkboxes, and the plan that should not get lost between long tests and parallel sessions.
Built in spare time: late nights, many full-day weekends at home, and ordinary windows around job, family, sport, and life.
Top priority · June 2026
Build the lab. Prove it on trading. Then attack ARC‑AGI‑3.
The order is deliberate: stabilize MAL v0.98 into v1.0, use the mature workflow for the Trading Arena v1.4 benchmark, and run ARC‑AGI‑3 as the first external Human × AI research challenge.
1 MAL2 Trading v1.43 ARC‑AGI‑3
v0.98 ready
Priority 2
Trading Arena v1.4
Use MAL as the prompt-building and audit layer, then test the upgraded real‑1m arena and BD Dual‑Side Size‑Parity Corridor Rescue without pruning high-risk strategies merely because a fixed wallet fails.
Priority 3 · external challenge
ARC‑AGI‑3 with MAL
Use HUMAN × AI brainstorming to build an agent that explores unfamiliar interactive environments, acquires goals, remembers discoveries, and adapts. Milestone #1 is a short sprint; Milestone #2 remains the fallback target.
ARC‑AGI‑3 command center
Everything needed before the next deadline
Official dates, registration, API setup, starter code, benchmarking tools, community links, and the offline-evaluation constraint in one place.
Calculating…Next ARC‑AGI‑3 milestone
Competition started
ARC‑AGI‑3 Milestone #1
ARC‑AGI‑3 Milestone #2
Submissions due
Papers due
Results announced
Immediate setup checklist
Windows quickstart
Platform login gives progress tracking and access to public games. Keep all secrets local.
pip install arc-agi
set ARC_API_KEY=your-api-key-here
python my-play.py
PowerShell alternative: $env:ARC_API_KEY="your-api-key-here". Follow the current official quickstart if commands change.
Official 2026 datesSchedule, goals, open-source rules
Kaggle competitionJoin, rules, code, submissions
ARC PlatformGoogle/GitHub login and scorecards
API key instructionsProfile → API Keys → ARC_API_KEY
Quickstart docsInstall toolkit and play first game
ARC‑AGI ToolkitOfficial Python integration
Agent templatesOfficial ARC‑AGI‑3 agent examples
Kaggle StarterSubmission-oriented starter repository
Benchmarking harnessRepeatable evaluations and scorecards
Games / tasksExplore available environments
Email updatesOfficial ARC Prize newsletter signup
DiscordCommunity and competition discussion
team@arcprize.orgOfficial foundation contact email
Submission reality: official competition eligibility requires reproducible open-source work submitted through Kaggle, and internet access is unavailable during Kaggle evaluation. An API-dependent research agent therefore needs an offline-compatible submission path rather than a last-minute wrapper.
Official links and dates verified 2026-06-18. The countdown uses calendar dates in UTC; always confirm the exact Kaggle cutoff time before submission.
News & announcements
Short public-facing updates and current work signals.
2026-07-19
8zip ReZIP and 8zipLab now have a dedicated evidence-first research page.
8zip & 8zipLab separates the standard-ZIP mainline from the experimental FPLB branch, explains the verified v0.2.1 result, the 9.07% FPLB3 exact-dedup migration proof, the active v0.2.2 run, the ongoing ReZIP 24-archive arena, realistic upside, caveats, and the next gates.
8zip research
8zip & 8zipLab separates the standard-ZIP mainline from the experimental FPLB branch, explains the verified v0.2.1 result, the 9.07% FPLB3 exact-dedup migration proof, the active v0.2.2 run, the ongoing ReZIP 24-archive arena, realistic upside, caveats, and the next gates.
2026-06-18
MAL v0.98 Human × AI is ready; ARC‑AGI‑3 becomes the next external challenge.
The new local workflow adds AI‑ONLY/HUMAN × AI, ASK_BD questions, BD seed injection, resume-safe ledgers, and contribution tracing. The immediate sequence is MAL self-loop → Trading Arena v1.4 benchmark → focused ARC‑AGI‑3 agent experiment.
Top priority
The new local workflow adds AI‑ONLY/HUMAN × AI, ASK_BD questions, BD seed injection, resume-safe ledgers, and contribution tracing. The immediate sequence is MAL self-loop → Trading Arena v1.4 benchmark → focused ARC‑AGI‑3 agent experiment.
2026-06-14
8Z Post-Codec Noise Scout — first master validation completed.
8Z Post-Codec Algorithmic Noise Scout now has a reviewed arena-results section. v0.1.1 recovered raw synthetic CA/PRNG structure with clean controls, but did not yet prove useful shrinkage on zlib/lzma final-byte compressed targets.
Noise Scout
8Z Post-Codec Algorithmic Noise Scout now has a reviewed arena-results section. v0.1.1 recovered raw synthetic CA/PRNG structure with clean controls, but did not yet prove useful shrinkage on zlib/lzma final-byte compressed targets.
2026-06-12
AIm³ MentalArena / mRHP — next reasoning layer page is staged.
AIm³ MentalArena defines mRHP: RHP improving RHP through protocol genes, concept capsules, benchmark tasks, scoring, lineage, reports, live extracts, and next-session seeds. It is an external workflow layer above LLMs, not model-weight learning and not a consciousness proof.
MentalArenaAI8 ReasoningAIm³ MentalArena defines mRHP: RHP improving RHP through protocol genes, concept capsules, benchmark tasks, scoring, lineage, reports, live extracts, and next-session seeds. It is an external workflow layer above LLMs, not model-weight learning and not a consciousness proof.
2026-06-12
How LLMs Actually Work — new RHPm / multi‑LLM explainer is live.
How LLMs Actually Work is now published as a user-facing, EN/SL article: transformer core, deployed assistant stack, failure modes, source discipline, and BD’s clearly separated AIM³ / RHPm / AI8 / 8Z workflow contribution. The companion scoreboard and comparison report keeps the multi‑LLM review trail separate from the main article.
LLM articleHow LLMs Actually Work is now published as a user-facing, EN/SL article: transformer core, deployed assistant stack, failure modes, source discipline, and BD’s clearly separated AIM³ / RHPm / AI8 / 8Z workflow contribution. The companion scoreboard and comparison report keeps the multi‑LLM review trail separate from the main article.
2026-06-11
NAS / Architecture Search Arena now has its own page.
NAS / Architecture Search Arena is now published as a dedicated MDL×DCC arena page, with Phase A structure discovery, Phase B controller/governance search, NB101 stress testing, and a downloadable code capsule.
NAS ArenaNAS / Architecture Search Arena is now published as a dedicated MDL×DCC arena page, with Phase A structure discovery, Phase B controller/governance search, NB101 stress testing, and a downloadable code capsule.
2026-06-09
Nara storytelling branch started.
Nara / Storytelling is now staged as a separate original-IP story branch under the future Art, Images, Music & Storytelling direction: power without domination, consent gates, visible cost, and the first episode, Ordinary Light. It is linked here in the BD companion hub while the root atlas stays unchanged until we decide to publish it wider.
NaraNara / Storytelling is now staged as a separate original-IP story branch under the future Art, Images, Music & Storytelling direction: power without domination, consent gates, visible cost, and the first episode, Ordinary Light. It is linked here in the BD companion hub while the root atlas stays unchanged until we decide to publish it wider.
2026-06-08
PiC page added to CRP.
PiC — π / Constants Compression is now linked as a normal CRP research page beside PiX: constants as deterministic MDL generator seeds for compression experiments.
PiCPiC — π / Constants Compression is now linked as a normal CRP research page beside PiX: constants as deterministic MDL generator seeds for compression experiments.
2026-06-07
Seed Bank renewal plan is live.
Seed Bank reframes MDLxDCC.org as a public memory graph for humans + future AI/ASI; Code Capsules defines the source/evidence standard.
Purpose layerSeed Bank reframes MDLxDCC.org as a public memory graph for humans + future AI/ASI; Code Capsules defines the source/evidence standard.
2026-06-06
Three new BD × AI Lab seed pages are live.
Physics × CFH × MDL×DCC frames hard physics questions through the method lens; Aliens on Earth? separates public UAP evidence, speculation, and resonance-test seeds; Talking to Aliens turns the contact question into a story-shaped entrance and civilizational mirror.
PhysicsAliens on EarthTalking to AliensPhysics × CFH × MDL×DCC frames hard physics questions through the method lens; Aliens on Earth? separates public UAP evidence, speculation, and resonance-test seeds; Talking to Aliens turns the contact question into a story-shaped entrance and civilizational mirror.
2026-06-06
Most important review discovery: code changed the scores.
Across review rounds, the biggest credibility shift happened when LLMs moved from reading HTML pages to inspecting Python code, working arenas, and real demo websites. That is now highlighted on the main atlas and review page.
ReviewAcross review rounds, the biggest credibility shift happened when LLMs moved from reading HTML pages to inspecting Python code, working arenas, and real demo websites. That is now highlighted on the main atlas and review page.
2026-06-05
PiX Universal Arena completed its first overnight run.
The first arena run completed 3,341 jobs with 10 workers and 0 failures. The next PiX v0.3 path moves toward real project folders and the trading-arena component.
PiX v0.3The first arena run completed 3,341 jobs with 10 workers and 0 failures. The next PiX v0.3 path moves toward real project folders and the trading-arena component.
2026-06-05
Current active focus clarified.
Main work now: website launch prep, trading arena with new PiX part, and TSP / 8Z-RP. Evidence packs and physical first-test plans stay as support lanes.
Active workMain work now: website launch prep, trading arena with new PiX part, and TSP / 8Z-RP. Evidence packs and physical first-test plans stay as support lanes.
Next
First public reproducibility pack.
Best next evidence move: start with 8Z-RP / TSP, then DNA/FASTA. Move from “LLMs saw the code” to “outsiders can run it.”
PacksBest next evidence move: start with 8Z-RP / TSP, then DNA/FASTA. Move from “LLMs saw the code” to “outsiders can run it.”
Broader work map
Support lanes beneath the top priorities
The top-priority sequence above governs immediate work. These wider lanes remain available in parallel while long tests run.
Lane 1
Website / public launch prep
Make the atlas clean enough before wider promotion.
Lane 2
Trading arena + PiX component
Broad strategy space first, including high-risk variants, then distill only what survives. PiX now belongs here as a controlled candidate-generator part of the trading/research arena.
Lane 3
TSP / 8Z-RP
Strongest technical proof lane and best arena-memory candidate.
Lane 4
Evidence Pack Factory
Turn code and results into reproducible public proof packs.
Lane 5
Physical first-test planning
Prepare cheap physical validation paths, especially HTH.
Saved locally
Todo checklist
Checkboxes and notes are saved in this browser via localStorage. Use Copy / Download / Email to preserve the plan outside the browser.
Lane 1 · Website / launch prep
Lane 2 · Trading arena + PiX component
Lane 3 · TSP / 8Z-RP
Lane 4 · Evidence Pack Factory
Lane 5 · Physical first-test planning
Backlog / incubator
Private plan notes
Saved in this browser only. Included in Copy / Download / Email plan.
Public credibility path
Evidence pack order
The fastest credibility gain is not more pages. It is clean reproduction packs.
1 · TSP / 8Z-RP
Strongest proof lane. Package one clean run with known instance, seed, config, expected result, and summary.
2 · Trading Clean API path
Distill survivors, then prove execution path through API paper, shadow live, and tiny live gates.
3 · 8Z / DNA / RouteSignal
Pick the most reproducible data-system pack after the first two, with manifest and one-command test.
Companion links
Quick access
Root stays the complete map. These are practical shortcuts for current work.
AIm³ MentalArena / mRHPNext reasoning layer: RHP that improves RHP through protocol genes, benchmark tasks, scoring, lineage, and next-session seeds.
/crp/AIM3_MentalArena_mRHP.html
How LLMs Actually WorkBeginner-first EN/SL explainer of base model, assistant stack, and the separated AIM³/RHP/AI8 workflow layer./BD/How_LLMs_Actually_Work.html
AI8 / 8Z ReasoningThe human + LLM invention method page that now connects to AIm³ MentalArena / mRHP./crp/AI8_Reasoning.html
8zip ReZIP & 8zipLabCurrent compression research: verified standard-ZIP optimization, experimental FPLB dictionaries, exact deduplication, live-run status, limits, realistic scenarios, and roadmap./8Z/8zip.html
Root AtlasFull public portfolio and crawler map./
Nara / StorytellingOriginal consciousness-centered story branch: power without domination, consent, cost, and ordinary light./Nara/index.html
APSVActive Pressure-Skin Vehicle: physical-device / active-skin arena page, with live-result notes and v0.3 seed backlog./crp/APSV.html
Seed Bank / Future AIPurpose layer and reading guide for humans + future AI/ASI./crp/SeedBank.html
AI ReviewsCurrent review synthesis for the portfolio./BD/BD_Idea_Atlas_AI_Reviews.html
llms.txtShort crawler/LLM map./llms.txt
llms-full.txtFull crawler/LLM map./llms-full.txt
Core ResearchMDL×DCC, arenas, devices, papers./crp/
NAS / Architecture Search ArenaDedicated MDL×DCC code-capsule page for neural architecture search and controller/governance experiments./crp/MDLxDCC-Arena-NAS.html
8Z Post-Codec Noise ScoutPhase-0 arena results: synthetic generator recovery, clean controls, no compressed-stream win yet, and the next bridge toward real codec/8Z residual layers./crp/8Z_PostCodec_Algorithmic_Noise_Scout.html
Physics × CFH × MDL×DCCHard physics questions through the CFH × MDL×DCC method lens./crp/Physics_CFH_MDLxDCC.html
Aliens on Earth?UAP evidence arena plus resonance-hypothesis seed paper./crp/AliensOnEarth.html
Talking to AliensTwenty alien questions as a story-shaped entrance into the wider BD × AI Lab material./crp/TalkingToAliens.html
AC / CCHConsciousness hypothesis stack./acp/
8ZCompression/data systems hub./8Z/
TradingTrading research branch./8ZT/
RHPAI brainstorming protocol./BD/BD_AIM3_RHP.html
RHPrPrompt/retrieval protocol./BD/BD_AIM3_RHPr.html
Trip OptimizerPractical route/trip demo./Trip/
ChessBestChess analysis / DCC demo./chess.html
Plan source notes
This public plan is a compact synthesis of the current conversation, AI8 operational/idea files, the 8Z TSP master plan, and the code-aware review notes. It intentionally avoids old internal review clutter and keeps only next useful moves.
- AI8O carries active execution state and next actions.
- AI8I carries seeds, bridges, and cheapest tests; not daily todo.
- AI8F carries compact continuity, not growing operational detail.
- TSP master plan defines JSON knowledge, transparency, Rust acceleration, challenger framework, gated MSTD-MR, GPU sensor numerics, and adaptive solver path.