MOTO v1.1.00#49
Closed
Intrafere wants to merge 10 commits into
Closed
Conversation
Prevent cleared autonomous sessions from resuming while preserving completed run history.
# MOTO v1.0.9 ## Features - Users can now utilize ChatGPT subscription through oAuth for inference, in replacement of or in combination with OpenRouter/LM studio. - Users can now do proof-only solving by disabling paper writing under the acceptable outputs control, this enables the existing brainstorming + Lean 4 automated proof solving loop without automated research paper writing. ## Changes - Updated proof verification prompts to prioritize direct user-prompt solution attempts and carry brainstorm topic/source-title context through theorem discovery, lemma search, SMT translation, and Lean formalization. - Persisted advanced runtime settings for proof concurrency, Lean/SMT options, and OpenRouter free-model fallback controls. - Lean 4 proof writing mode has stronger emphasis on seeking novelty. - Cleaned up .cursor rules and aligned them with program where outdated. - Removed old/outdated code and scaffolding. - Improved live activity GUI logging. - Fresh starts now open on the main Autonomous screen instead of restoring the last viewed screen. - Cleaned up main directory. - Topic selection brainstorming sessions now emphasize to aggressively pick topics that solve the user's whole prompt at once where possible. - Expanded GUI live activity logging before truncation. ## Bug Fixes - Proof-solving attempts now receive the complete source paper or brainstorm context, making all models much more effective at solving Lean proof targets. Prior to this fix, all proof attempts were missing the accompanying brainstorm/paper which meant there was no Top-P exploratin architecture assisting proof solving. This is why only SOTA models were able to solve proofs before. - Hardened diagnostic logging against CodeQL log-injection findings and cleaned up first-party CodeQL warning-level frontend/backend issues. - Performed CodeQL audit and made minor fixes. - Added dedupe script for duplicate proof context, greatly reducing duplicate proof context. - Removed hidden context/max-token fallbacks and fixed compiler RAG budgeting so LM Studio, OpenRouter, and Codex use user-configured limits. - Fixed compiler High Parameter rigor proofs so Lean-verified proofs are ranked and indexed under the active paper. Authored by Patrick White, Patrick@Intrafere.com
# MOTO v1.0.9 ## Features - Users can now utilize ChatGPT subscription through oAuth for inference, in replacement of or in combination with OpenRouter/LM studio. - Users can now do proof-only solving by disabling paper writing under the acceptable outputs control, this enables the existing brainstorming + Lean 4 automated proof solving loop without automated research paper writing. ## Changes - Updated proof verification prompts to prioritize direct user-prompt solution attempts and carry brainstorm topic/source-title context through theorem discovery, lemma search, SMT translation, and Lean formalization. - Persisted advanced runtime settings for proof concurrency, Lean/SMT options, and OpenRouter free-model fallback controls. - Lean 4 proof writing mode has stronger emphasis on seeking novelty. - Cleaned up .cursor rules and aligned them with program where outdated. - Removed old/outdated code and scaffolding. - Improved live activity GUI logging. - Fresh starts now open on the main Autonomous screen instead of restoring the last viewed screen. - Cleaned up main directory. - Topic selection brainstorming sessions now emphasize to aggressively pick topics that solve the user's whole prompt at once where possible. - Expanded GUI live activity logging before truncation. ## Bug Fixes - Proof-solving attempts now receive the complete source paper or brainstorm context, making all models much more effective at solving Lean proof targets. Prior to this fix, all proof attempts were missing the accompanying brainstorm/paper which meant there was no Top-P exploratin architecture assisting proof solving. This is why only SOTA models were able to solve proofs before. - Hardened diagnostic logging against CodeQL log-injection findings and cleaned up first-party CodeQL warning-level frontend/backend issues. - Performed CodeQL audit and made minor fixes. - Added dedupe script for duplicate proof context, greatly reducing duplicate proof context. - Removed hidden context/max-token fallbacks and fixed compiler RAG budgeting so LM Studio, OpenRouter, and Codex use user-configured limits. - Fixed compiler High Parameter rigor proofs so Lean-verified proofs are ranked and indexed under the active paper. Authored by Patrick White, Patrick@Intrafere.com
Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
## Features ## Changes - Autonomous brainstorm and paper proof checkpoints now run up to four strict prompt-solving follow-up rounds. - oAuth is now a valid startup choice when doing the new user startup. - Added "try to prove this" button with Lean 4 prover to manual/advanced mode. - Cleaned up confusing wording on proof novelty review from, removed "mechanized proof is novel for the research program" wording to clarify proof must be novel for society, not just the research program. - "Try to prove this" button now appends solved proofs to the brainstorm regardless of novelty, allowing for a better Ralph loop in iterative uses. ## Bug Fixes - Fixed proof-check source assembly so manual and autonomous "Try to Prove This" paths strip generated proof appendices, inject verified proofs only through the proof library, and keep proof candidates focused on directly solving or building toward the user prompt. - Fixed saved manual compiler paper proof checks to include the full proof-stripped manual Aggregator context instead of a hidden truncated excerpt. - Fixed fresh GitHub ZIP installs falsely showing an update-available notification on first launch. - Fixed stale Aggregator rejection feedback leaking across manual clears and autonomous child aggregators. - Fixed non-Windows Lean/elan bootstrap to use the live upstream GitHub installer script instead of the retired lean-lang.org URL. - Fixed ChatGPT/Codex oAuth retry bug causing failed proofs. - Fixed Manual Writer proof output isolation with a separate Mathematical Proofs tab/store so it no longer pollutes Autonomous Research proofs. Authored by Patrick White, Patrick@Intrafere.com
## Features - Grok/SuperGrok OAuth is now enabled for inference use within the harness. ## Changes - Leaderboard updated: 1st place - Kimi K2.6, 2nd place - ChatGPT 5.5, 3rd - Autonomous brainstorm and paper proof checkpoints now run up to four strict prompt-solving follow-up rounds. - Added "try to prove this" button with Lean 4 prover to manual/advanced mode. - Cleaned up confusing wording on proof novelty review from, removed "mechanized proof is novel for the research program" wording to clarify proof must be novel for society, not just the research program. - "Try to prove this" button now appends solved proofs to the brainstorm regardless of novelty, allowing for a better Ralph loop in iterative uses. - History prompt previews now truncate to 400 characters. - Manual proof checks now use standard multi-candidate proof discovery instead of a direct Lean target shortcut. - OpenAI Codex no longer has a hard coded autofill fallback for models when the server fails to return a model list. - OpenAI Codex now handles token revoked errors better. - Removed "fast, affordable, mid-tier knowledge" profile. Updated Slow, affordable, high knowledge profile. - Lean 4 startup status now uses softer "still starting up" language and reduced polling/log spam. ## Bug Fixes - Fixed proof-check source assembly so manual and autonomous "Try to Prove This" paths strip generated proof appendices, inject verified proofs only through the proof library, and keep proof candidates focused on directly solving or building toward the user prompt. - Fixed saved manual compiler paper proof checks to include the full proof-stripped manual Aggregator context instead of a hidden truncated excerpt. - Fixed fresh GitHub ZIP installs falsely showing an update-available notification on first launch. - Fixed stale Aggregator rejection feedback leaking across manual clears and autonomous child aggregators. - Fixed non-Windows Lean/elan bootstrap to use the live upstream GitHub installer script instead of the retired lean-lang.org URL. - Fixed ChatGPT/Codex OAuth retry bug causing failed proofs. - Fixed Manual Writer proof output isolation with a separate Mathematical Proofs tab/store so it no longer pollutes Autonomous Research proofs. - Manual mode validator setting no longer falls back to autonomous mode validator setting when no option is chosen without warning. - Fixed React refresh bug causing the settings drop down to close on the user. Authored by Patrick White, Patrick@Intrafere.com
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
MOTO v1.1.00
Features
Changes
research program" wording to clarify proof must be novel for society, not just the research program.
Bug Fixes
Fixed proof-check source assembly so manual and autonomous "Try to Prove This" paths strip generated proof appendices, inject verified proofs only through the proof library, and keep proof candidates focused on directly solving or building toward the user prompt.
Fixed saved manual compiler paper proof checks to include the full proof-stripped manual Aggregator context instead of a hidden truncated excerpt.
Fixed fresh GitHub ZIP installs falsely showing an update-available notification on first launch.
Fixed stale Aggregator rejection feedback leaking across manual clears and autonomous child aggregators.
Fixed non-Windows Lean/elan bootstrap to use the live upstream GitHub installer script instead of the retired lean-lang.org URL.
Fixed ChatGPT/Codex OAuth retry bug causing failed proofs.
Fixed Manual Writer proof output isolation with a separate Mathematical Proofs tab/store so it no longer pollutes Autonomous Research proofs.
Manual mode validator setting no longer falls back to autonomous mode validator setting when no option is chosen without warning.
Fixed React refresh bug causing the settings drop down to close on the user.
Authored by Patrick White, Patrick@Intrafere.com