Selected proposals from my sustained use of AI for research, professional development and knowledge management. These recommendations address accessibility, transparency, user control and the practical work of preserving evidence.
These are examples from many proposals and feedback submissions. Each entry gives the requested improvement in brief; the response record can be expanded.
· GitHub feature request · Published, not implemented
Codex: automated relevance and safety approval, AI prototyping, human verification and release
Requested a structured workflow separating initial relevance/reasonableness assessment from comprehensive safety review, followed by sandboxed AI-led development and automated tests. Human staff would verify predefined acceptance criteria, while controlled release would occur only after required authorisations.
Public record:openai/codex #59267 (verified as open and labelled enhancement at inspection). No product adoption or engineering decision is claimed.
Hugging Face: Claim-to-Evidence Coverage & Contradiction API
Published a Hugging Face forum proposal for sub-claim evidence mapping, contradiction statuses and provenance. Related diagrams are available below; technical review and adoption are unconfirmed.
Submitted IBM Idea APICONN-I-1245, proposing evidence challenges, appeal decisions and versioned correction propagation. Submission link supplied; technical review and implementation are unconfirmed.
Progress recorded:
Portable, consent-based evidence verification
Developed a cross-platform proposal for retrieving reviewed findings from public and authorised non-public evidence. Defined reviewer and provider responsibilities, company and user benefits, a ten-step data flow, correction procedures and privacy safeguards.
Outreach and status
Extended outreach to deepset, LlamaIndex, Jina AI, Sakana AI, MiniMax and Z.ai. Six individual sends were confirmed by Gmail and retained under the AI Platform Proposals archive label. No immediate bounce was found at the sending check; delivery, substantive review and adoption remain unconfirmed.
This is a proposal and correspondence record, not an operating verification network or certification of personal claims. Private emails and source documents are not published here.
External GitHub discussion reference · site record
Mistral Vibe — GitHub Discussion #1393
Public Mistral Vibe discussion supplied as an additional record in the AI feature-proposal and contributor-outreach register. Its exact discussion title, full text, author and original posting date have not been independently verified here; the site record date is not presented as the GitHub publication date. Mistral Vibe discussions are product-specific and do not imply company-wide consideration or adoption.
Contributor Provenance and Innovation Attribution Records — Hugging Face Community
Public proposal concerning traceable records of intellectual and technical contributions, including distinctions between human-originated conceptual work, collaborative development and AI-assisted articulation. Recorded as a platform-specific public contribution; this summary describes the topic, not an independently verified reproduction of the complete post.
Attribution: Grace R. Nettey — Manifold Grace. Submission and publication do not establish technical acceptance, adoption, intellectual-property ownership or implementation.
Contributor Provenance and Innovation Attribution Records — OpenAI Developer Community
Public proposal concerning records of contributor provenance and innovation attribution. It relates to documenting identifiable conceptual contributions and distinguishing human-led problem framing and ideas from AI-assisted drafting or expression. The full posted text could not be independently retrieved for this record; this summary reflects its stated topic and the associated research context.
Attribution: Grace R. Nettey — Manifold Grace. Publication of a proposal does not establish independent novelty, institutional endorsement, review or implementation.
Historical As-of-Date Evidence — Google AI Developers Forum (Gemini)
A public feature request concerning reliable historical evidence-state retrieval: differentiating the date of an event, the time a record was published or recorded, its status as of a requested date, and subsequent amendments. Linked as a related platform-specific submission; the precise published forum text has not been independently retrieved for this site record.
Proposal attribution: Grace R. Nettey — Manifold Grace. Human-led conceptual research with AI-assisted structuring and drafting. Public posting does not establish review, adoption or implementation.
Historical As-of-Date Evidence Retrieval API — AWS Builder Wishlist
Proposed an optional as_of parameter for historical evidence retrieval, with record version, publication and effective dates, historical verification status, provenance, and separate notices of later corrections. The proposal distinguishes when events happened from when evidence was published or recorded, and does not assume historical records were accessible to everyone.
Attribution: Grace R. Nettey — Manifold Grace. Human-led conceptual research with AI assistance in structuring and drafting. Published as a feature request; review, adoption and implementation are not established.
Conversation transparency and knowledge management
Measure conversation histories and show what information is stored, retrieved, active or summarised, with topic maps, chronology and structured exports.
Response record
Support acknowledged the proposal and said its key themes were captured for product feedback review.
13 May
Professional profile, CV and portfolio synchronisation
Maintain structured career records and turn research, chats and projects into consistent CV and portfolio updates across chosen platforms. Publishing would remain user-controlled, with previews, granular permissions, version history and rollback.
Response record
AI-assisted support acknowledged the detailed proposal and said it captured the request for product feedback; formal internal logging was not confirmed.
Response: 3 June
Advanced research workspace
Provide an integrated workspace for research, evidence logs and knowledge management, supporting sustained work across complex projects.
Response record
Support recognised the value of the suggestion and said it would raise the feedback with engineering for evaluation.
Response: 3 June
Message timestamps and audit logs
Provide exact message and file-upload timestamps, session duration and exportable audit logs for research, accessible record-keeping and governance.
Response record
Support acknowledged the potential value and said it would pass the feedback to the team for consideration.
Response: 3 June
Priority folders and archived-chat organisation
Organise archived conversations with folders, categories, tags, priorities, project groupings, summaries and better search.
Response record
Support acknowledged the potential value for long-running projects and said it would pass the feedback to the team for consideration.
Response: 8 October 2026
Read-only connections and source integrity
Offer a dedicated read-only Google Drive connection, with optional write access, permission upgrades by choice and a clearly visible access level.
Response record
Support acknowledged the detailed recommendations and screenshots; implementation remains unconfirmed.
2 November 2026 · Disputed moderation outcome
NVIDIA Developer Forums: Evidence-Gap Discovery & Referral API
Proposal: Identify material gaps in available evidence and suggest proportionate, appropriately authorised requests to qualified experts or institutions; preserve source provenance, uncertainty, decisions and human oversight. Potential uses include research integrity, public administration, safeguarding and authorised investigative support. These are proposed benefits, not validated deployments.
Recorded response: A NVIDIA Developer Forums notification dated 2 November 2026 reported deletion of the contributor's account following classification of a submission as promotional spam. The contributor disputes that classification, describes this as their only forum submission and maintains that the purpose was independent technical research rather than commercial advertising. Screenshots also document a forum login error and separate NVIDIA registration/access errors; a common cause is not established.
Evidence still needed: Original submission and URL, forum category, exact wording and links, rule relied upon, moderation rationale, review procedure and outcome. The notification does not establish that NVIDIA's engineering team assessed or rejected the proposed architecture, or why the moderator reached the classification.
Status: Disputed moderation; technical review and implementation unconfirmed. The contributor has prepared requests for preservation of records, explanation and independent reconsideration. The wider impact of the decision has not been measured.
20-company feature-gap scan → — compare documented boundaries, existing functionality, possible research tests and appropriate feedback routes.
Historical cross-platform feature and contribution register
Earlier ideas, submissions outside AI platforms, dated outreach and recipient feedback are consolidated in the Feature Requests & Contribution Register. The register separates original concepts, repeated outreach, acknowledgements, moderation, unverified responses and adoption; existing individual proposal records remain here.
Follow-on research, distinct from original submissions
The 4 November 2026 synthesis extends earlier feature requests into a proposed evaluation architecture. Original submission dates, wording and evidence are not replaced. Read the proposed research framework.
Additive research note, 4 November 2026. Conceptual and testable; not a validated ASI capability.