AI Longevity Project Opportunities

Bottom Line Up Front (Use This to Prioritize Your Work)

• CosinorAge ( https://github.com/ADAMMA-CDHI-ETH-Zurich/CosinorAge ): A Python package from ETH Zurich’s ADAMMA-CDHI lab, self-described as implementing Cosinor model-based biological age prediction. Whether it’s runnable today is TBD — license, data entry points, and dependencies weren’t included in the source materials. Great fit for a Python tools directory entry and a test-run writeup; not for personal biological age assessment or health advice. A self-described README ≠ clinical validation.

• epiage-skill ( https://github.com/gangchen/epiage-skill ): Self-described as containing 24 epigenetic clock algorithms, only depends on pandas + numpy, claims to run offline. License and data entry points still need verification. Good for indie devs who want to tear it apart; not for scoring personal methylation data or making health/risk judgments.

Today’s Priority Projects 🎯

CosinorAge

CosinorAge is an ETH Zurich-backed Python package that explicitly states “biological age prediction” in its description, with an active signal as recent as 2026-10-25. It fits right into the longevity bio-age estimation tools ecosystem — solid candidate for a tools directory entry and test-run notes. Specific input data types and whether example data is bundled still need to be confirmed in the README.

• Evidence source: https://github.com/ADAMMA-CDHI-ETH-Zurich/CosinorAge (GitHub Trending, last active 2026-10-25T11:46:19Z)
• Confidence level: Medium (institutional repo, clear description, but README, license, and data entry points weren’t in the source materials)
• Target audience: Indie devs, longevity content creators, research learners
• What problem it solves: Uses the Cosinor model to extract parameters from rhythmic signals and estimate biological age — specific input data types still TBD
• Test-run barrier: Medium — Python environment required, but input data format and example datasets still need verification
• License & data: TBD
• Maintenance overhead: Medium — institutionally maintained, but you’ll need to track upstream updates
• What you can make with it: Test-run tutorial / tools directory entry / resource library entry
• Deliverables: README test-run notes, install screenshot, dependency list, 3 key takeaways
• Minimum viable validation: Open the README and check whether install steps and example data exist
• Success metric: Complete pip install and locate example input data within 24 hours
• Stop conditions: License prohibits commercial use; no example data and data format docs are missing; dependency conflicts can’t be resolved
• Secondary dev paths: ① Add as an entry in a longevity Python tools directory page; ② Write an algorithm breakdown tutorial; ③ Build it into a tools radar entry
• Post-publish / compliance risk: Medium
• Compliance boundaries: Frame everything as “research tool observation” and “algorithm breakdown” — do not describe it as a personal biological age measurement service or imply any anti-aging effects
• Today’s minimum action: Open the repo, check the README, license file, and examples/ directory — screenshot it and jot down 3 key takeaways

Secondary Development Directions 🛠️

• Direction A: epiage-skill algorithm inventory. epiage-skill ( https://github.com/gangchen/epiage-skill ) self-describes as containing 24 epigenetic clocks including GrimAge V1/V2, Horvath, and PhenoAge. Deliverable: A draft comparison table header mapping algorithm names × data types. Minimum viable validation: Open the repo today and verify whether the license and README algorithm list actually match the description. Upgrade path: If the license allows, turn it into a lightweight nav page or a recurring newsletter section — “Epigenetic Clock Radar.”

• Direction B: NeuroXT FDA-cleared tool regulatory timeline. NeuroXT NEXT-Cue ( https://longevity.technology/news/fda-clears-neuroxts-ai-tool-that-sizes-up-alzheimers-risk/ ) received FDA clearance, pairing brain MRI with cognitive scoring for clinicians. Deliverable: One resource library entry covering the tool name, use case (clinical, not consumer), regulatory status, and source link. Minimum viable validation: Bookmark the news URL and draft one content brief with the explicit note “clinical use only, not a personal diagnostic tool.” Upgrade path: Track similar FDA-cleared AI neurocognitive tools and build a monitoring page.

Worth Watching 👀

• brain-age-prediction-ixi ( https://github.com/anupbagale/brain-age-prediction-ixi ): 3D CNN for brain age prediction using the public IXI dataset, built in Jupyter Notebook, last active 2026-10-30. License and dependencies still TBD; compute requirements for 3D MRI processing are unclear. Why we’re holding off: Compute and license unconfirmed — waiting to see if any test-run records surface before evaluating.

• Mayo Clinic precure research ( https://www.news-medical.net/news/20261030/Mayo-Clinic-enters-new-medical-era-with-precure-research.aspx ): Pre-symptomatic disease biomarker research direction, currently just a press release — no specific paper or tool link. Why we’re holding off: No actionable project leads yet. Will revisit when a concrete paper or tool drops.

Don’t Touch Today 🚫

• Butterfly_DNAm_age ( https://github.com/JesperBoman/Butterfly_DNAm_age ): Self-described as a butterfly epigenetic clock training script based on animal data, 1 star, written in R, license and data TBD. Animal research doesn’t apply to personal health assessment, activity is low, and there’s no content repurposing value right now.

• Rivus RV-8451 ( https://longevity.technology/news/rivus-to-show-animal-data-on-muscle-sparing-glp-1/ ): Preclinical animal trial — results will be presented as a poster at ObesityWeek in November, no published paper yet. Animal trial findings should not be repurposed into any personal health judgment tool or intervention recommendation, period.

Today’s Action Items ✅

• First test-run: Open https://github.com/ADAMMA-CDHI-ETH-Zurich/CosinorAge , check the README and license file, and note whether install steps and example data exist
• First write: Draft one content brief — “ETH Zurich’s Biological Age Python Tool: How the Cosinor Model Works” (label it as a research tool observation, not a personal measurement service)
• First bookmark: https://github.com/gangchen/epiage-skill — for the upcoming algorithm comparison table
• First table draft: After checking the epiage-skill README, sketch out an algorithm name × data type table header (contents based on actual README, not assumptions)
• First avoid: Animal experiment projects (Butterfly_DNAm_age, Rivus RV-8451) and the Mayo precure news with no public data — and publish nothing that resembles a personal biological age assessment or cognitive risk judgment