I built a crawler that finds each hospital's official file and pulls out 29 common services. Across 482 hospitals: a moderate ER visit is listed at a median $1,285 while insurers pay $288; a metabolic panel is listed at $207 while insurers pay $11.46.

To keep it honest: only per-service fee-schedule dollar rates count (no case rates or percentages, which cover a whole visit), each figure needs at least 3 insurers, and every value is checked against Medicare reference rates. Surgery is excluded because insurer surgery rates cover the whole operation while the list price is one line item. Each hospital page links its source file. Feedback on the methodology is very welcome.

Comments URL: https://news.ycombinator.com/item?id=49926185

Points: 32

# Comments: 15

TokenShield is a small Node 22 proxy that sits in front of api.anthropic.com. Set `ANTHROPIC_BASE_URL=http://127.0.0.1:7777` and Claude Code (or Cursor, Windsurf, Zed, Aider — anything that uses the Anthropic SDK) routes through it. The proxy hashes every tool_result content block per-conversation; subsequent occurrences of the same hash within the same conversation are replaced with a deterministic stub:

  { "type": "tool_result",
    "tool_use_id": "...",
    "content": "[tokenshield: identical content seen at message 2, sha:1f6063fe]" }

The accounting side was the bug-hunt I didn't expect. First version captured 0 tokens on successful responses. Turned out Anthropic returns Content-Encoding: gzip and I was running JSON.parse on the compressed buffer. The silent try/catch swallowed the error and I shipped "200 status, $0.00 spent" for a week before noticing. Fix: `accept-encoding: identity` on upstream — bandwidth on localhost doesn't matter and reliable measurement does.

Real numbers from my own usage yesterday: one Opus-4 turn paid $0.1719, saved $0.0747 — 30.3% on that turn. Bench numbers across three recorded fixtures (light Q&A, medium coding, heavy 25-turn agentic): 0%, 27.7%, 62.1%. The light case correctly does nothing — there's nothing to dedup in a 5-turn conversation with no tool use.

What's deliberately not here:

- It only helps API-billed clients. Claude Desktop and claude.ai web are flat-subscription / hardcoded endpoint — no metered bill to compress. - OpenAI / Gemini are on a waitlist. The provider adapter interface is in the code, but I'd rather ship one provider correctly than three half-baked. - Compression is conservative (>= 256 byte payloads, content-hash keyed). It does not touch streaming SSE bytes — passthrough is byte-faithful so the prompt cache stays valid.

The piece I'd most appreciate feedback on: I'm staring at `response-cache` as the next processor (LRU+TTL on (model, system_prompt_hash, last_user_msg_hash)), but the correctness story scares me — `temperature > 0` and tool-using conversations both make caching risky. Would love thoughts from anyone who has done semantic caching on LLM responses about where the landmines are.

`npm i -g @curatedmcp/tokenshield` to try it. MIT, self-hosted, your API key never leaves the machine, telemetry is opt-in and aggregate-only.

More at: https://curatedmcp.com/tokenshield

Comments URL: https://news.ycombinator.com/item?id=48197361

Points: 1

# Comments: 1

Comments URL: https://news.ycombinator.com/item?id=48180942

Points: 3

# Comments: 3