snow_mac

I wanted to query the entire archive of a podcast I listen to without paying for their coaching tier. Built a lightweight pipeline to ingest, summarize, index, and answer questions with citations back to the source audio.

Totals:

- 230 episodes (~145 hours of audio, 1.29M words)

- Total cost: ~$33 using OpenAI's transcription and embedding APIs

- Pipeline: Audio chunking -> Whisper transcription -> Structured summaries -> Semantic index -> Retrieval with source citations

A few engineering details from the post-mortem:

1. Concurrency traps: Initially went with unbounded async calls, which quickly triggered rate-limit stalls and hung threads. Throttled sequential batches ended up being both faster and predictable.

2. Silent failure modes: A shadowed `OPENAI_API_KEY` environment variable silently broke auth midway through a long run without failing fast.

3. Model swapping: Swapping transcription models mid-stream cut processing costs noticeably without hurting search recall.

Question for the HN crowd on retrieval: Right now, retrieval is bare-metal flat NumPy cosine similarity across ~1,500 chunk embeddings. It’s instantaneous in memory (<2ms) and avoids the overhead of running a dedicated vector DB.

For those who have pushed flat in-memory search further: at what order of magnitude (15k? 50k?) does this genuinely fall over before it makes sense to bring in something like `sqlite-vec`, `pgvector`, or an ANN index?

Maybe your argument works for a few specific cases, what about those that do not meet that? What about someone with stage 4 cancer who needs radiation and will likely live a year or two longer without treatment? Without treatment they can't recover, without recovering they can't go to work; their body is literally dying.

With treatment they might be able to live, work and recover. That is what the article is saying. Without insurance people die from cancer. With insurance some people do not die. If we cut their insurance due to more red tape, we're condemning those to death.

Evil. Just evil. Terminally ill patients can't work, they're on deaths door. If they don't get treatment, they are dead. Let's let the medical decisions be handled by doctors and not some politician.

Honestly things like this show me the President has to much power. There should be government agencies where no political party can touch it without significant oversight and approval from both the president and both houses of congress.

Medicare, Medicaid, FBI, FCC, FAA, The Fed, and more all should be outside admin control of the president and any single party or group of individuals. Only when you have a 2/3s majority of both the houses and the president should you be able to make such broad sweeping policy changes.

I am actively developing on 4 projects fulltime. Kind of like double or triple dipping but just 1 employer. With a Claude Code license I'm running 2-4 CLI instances with AI. I'm botsitting 6-8 hours a day and getting more done then every. I love how much I can build using AI

Created two days ago.... Did you use Claude Code to generate the open source code? :P

I love this. I recently found a few solar panels dumpster diving that put out a good load. Just basic scratches and that is all. I am building out my own system to power my house and in the next 2 years will be energy off grid in the city.

This is a good thing, the next step however is to block the guys that use sites like BiggerPockets.com from buying up more then 5 single family homes. I know a whole bunch of guys who own 10 or more houses. That is excessive. We need to limit all investment of single family housing. Apartment complexes? Be my guest buy all you want, build all you want. But lets keep our neighborhoods full of people who can actually live in them.

How do you get access get all the local government meetings? Do you have a crawler that looks up every city in the country then visits each website and pull down the info? A public listing site?

This is amazing. I wonder how it works. I would be cool if they published it.