ride-recap, teaching a LLM my taste to automate cycling highlights
TL;DR: Turn hours of raw GoPro footage + a .fit file into a 60-second highlight reel with ride telemetry burned in. Every second of the ride is scanned by gemini-3.5-flash, as is the clip ranking + curation. The whole thing costs about $0.04 per ride and takes 10 minutes. Longer version: Road cycling has been my primary form of exercise, social outlet, therapy, wardrobe expense, and personality trait for almost a decade. I ride most weekends, usually out of Manhattan and up 9W. By the end of a ride I have hours of GoPro footage and one .fit file with per-second speed, power, heart rate,…
What it does
In the maker’s words, at launch
TL;DR: Turn hours of raw GoPro footage + a .fit file into a 60-second highlight reel with ride telemetry burned in. Every second of the ride is scanned by gemini-3.5-flash, as is the clip ranking + curation. The whole thing costs about $0.04 per ride and takes 10 minutes. Longer version: Road cycling has been my primary form of exercise, social outlet, therapy, wardrobe expense, and personality trait for almost a decade. I ride most weekends, usually out of Manhattan and up 9W. By the end of a ride I have hours of GoPro footage and one .fit file with per-second speed, power, heart rate, cadence, and GPS. Absolutely no one wants to watch 3 hours of being stuck behind Citibikes on West Side Highway. It's fun to look through past footage, identify the fun parts, and put together a narrative to remember. But since cycling is already time consuming, manually editing a highlight reel edit per ride is a nonstarter. So I built and open-sourced https://github.com/ianmacomber/ride-recap. I identify compelling moments from four separate sources: * Garmin telemetry via .fit file (speed, HR, power spikes, sprints, climbs) * Strava via API (popular segments) * Gemini vision scan + rating of individual frames * (optional) hand-labels via Streamlit app The fusion step has a LLM narrative pass pick 20 clips to best tell the story of the ride, boosting “cross-source agreements” (if a human label + telemetry + Strava + Gemini all agree that a clip is interesting), with greedy re-ranking and a crowding penalty to avoid clips too close to something already selected. Obvious in retrospect, but there’s no substitute other than looking at the clips Gemini selects, being highly opinionated about what should / should not be included, being specific enough about why, and repeating until you can’t think of anything more to improve. You cannot teach an LLM taste if you do not have taste yourself.
Does the same job
all alternatives →- GCGemini can now natively embed video, so I built sub-second video searchMar 2026 · github.com · ▲438
Gemini Embedding 2 can project raw video directly into a 768-dimensional vector space alongside text. No transcription, no frame captioning, no intermediate text. A query like "green car cutting me off" is directly comparable to a 30-second video clip at the vector level. I used this to build a CLI that indexes hours of footage into ChromaDB, then searches it with natural language and auto-trims the matching clip. Demo video on the GitHub README. Indexing costs ~$2.50/hr of footage. Still-frame detection skips idle chunks, so security camera / sentry mode footage is much cheaper.
- APA Pipeline for Making 10-minute AI Movies with Claude Code and SeedanceJul 2026 · github.com · ▲21
I've been working on this Claude Code pipeline for making 10-minute movies, gluing together Seedance (videogen), Nano Banana (image gen), and ElevenLabs (voice gen), and using Claude as the director. My repo contributes a markdown playbook and a worked example which should make it easier for other people to make their own short movies. A first-pass movie takes around 2.5 hours of wall clock time and costs around $200, and iterating is easy thanks to intermediate artifacts like storyboards and voice samples.
More ai this month
the category →
I trained a 125M-parameter transformer to autocomplete piano performances in real time (~108 notes/sec on an iPhone 15). The idea is basically GitHub Copilot or Tabnine, except instead of prompting it with code, you prompt it by playing a few notes on a MIDI piano. The model then continues what you played, entirely on-device. The app is free if anyone wants to try it. Happy to answer questions about the model, training, Core ML, or the many things that didn't work.
AI · 17d ago · simedw.com
Astute▲585Automate your B2B brand going viral, with new media creators
AI · 18d ago · company-app.joinastute.com
Hey HN, Henry from Cactus here! We previously released Cactus Needle, a 14MB agentic LLM for tool call, device use, and structured extraction for phones, wearables, smart homes, small robots and microcontrollers. We got really great feedback here, and have now incorporated the suggestions to release Needle 2. The whole model is a single 14MB binary that runs a full session in 28MB of RAM; 45m parameters at 2bit compression. Needle hits 500 tokens/sec decode speed on a Raspberry Pi 5, sits between 400-1,500 tokens/sec on VR devices like Meta Quest 3S and Apple Vision Pro, and ranges…
AI · 27d ago · cactuscompute.com
Launched alongside, July 2026
the whole month →- IR
I might be the only SRE on Earth with his own bowling center. It's a more in-depth gig than you'd think. My family and I bought an abandoned 8-lane bowling center in the rural mid-west. In our small town there weren't many recreation options for families. You've heard of a food desert? This is an R&R desert. It had been abandoned for a good reason. The roof leaks, the electrical system was constantly surging, and my 70-year-old bowling equipment (still) doesn't work perfectly. The system that keeps your score is particularly interesting to me. It's the thing you watch during your game, but…
Life & fun · Jul 2026
- EElevators▲1,680
Life & fun · Jul 2026 · john.fun
- 1W18 Words▲1,160
Life & fun · Jul 2026 · 18words.com
- BA
Over the past few months, our team has been building more and more slidedecks using web frontend technologies with coding harnesses like Claude Code, but a common complaint is to make even small edits we need to edit the code either manually or via the harness. To avoid this loop, I ended up creating Bento, a single HTML file with everything you need in a slide tool including animations and shared editing. There's no install or cloud login, everything works offline. The default deck is around 560 KB and it doesn't need to fetch anything once you got it. Open it in a browser and then you can…
Dev tools · Jul 2026 · bento.page
- GG
A few days ago I found myself trying out GLM 5.2 and was really positively impressed. The capabilities and security I was getting from this LLM are similar to those I've gotten from models like Claude or GPT, and this really surprised me. But then I thought, "I wonder how it would work on a normal computer like mine," and above all, "I wonder if it would work without going into OOM on a computer like mine." So I started working with the help of agents to test this possibility. I started converting the model to int4, understanding MTP usage, and if possible implementing DSA for long context.…
AI · Jul 2026 · github.com
