Feed aggregator

Article URL: https://100jumps.org/play/

Comments URL: https://news.ycombinator.com/item?id=47361797

Points: 2

# Comments: 1

Article URL: https://old.reddit.com/r/dataisbeautiful/comments/1rs2wzq/oc_black_logos_are_taking_over_silicon_valley/

Comments URL: https://news.ycombinator.com/item?id=47361781

Points: 2

# Comments: 0

Article URL: https://www.technologyreview.com/2026/03/12/1134243/defense-official-military-use-ai-chatbots-targeting-decisions/

Comments URL: https://news.ycombinator.com/item?id=47361776

Points: 1

# Comments: 0

I built CacheLens because I was burning through $200+/month on Claude API calls and had no idea where it was going.

It's a local HTTP proxy that sits between your app and the AI provider (Anthropic, OpenAI, Google). Every request flows through it, and it records token usage, cost, cache hit rates, latency — everything. Then there's a dashboard to visualize it all.

What makes it different from just checking your provider dashboard:

It's real-time (WebSocket live feed of every call as it happens) It works across all three major providers in one view It runs 100% locally — your prompts never leave your machine It has budget caps that actually block requests before you overspend It identifies optimization opportunities (cache misses, model downgrades, repeated prompts) Tech stack: Python, FastAPI, SQLite, vanilla JS. No React, no build step, no external dependencies beyond pip. The whole thing is ~3K lines of Python.

Interesting technical decisions:

The proxy captures streaming responses without buffering — it tees the byte stream so the client sees zero added latency Cost calculation uses a built-in pricing table with override support (providers change rates constantly) There's a Prometheus /metrics endpoint so you can plug it into existing monitoring Cacheability analysis uses diff-based detection across multiple API calls to identify what's actually static vs dynamic in your prompts Limitations I'm honest about:

The cacheability scorer is heuristic-based — solid for multi-call traces (~85% accurate), rougher for single prompts (~65%) Token counting uses cl100k_base for everything, which drifts ~10% for non-OpenAI models Three features (smart routing, scheduled reports, multi-user auth) are on the roadmap but not shipped yet Would love feedback, especially from anyone managing LLM costs at scale.

Comments URL: https://news.ycombinator.com/item?id=47361756

Points: 1

# Comments: 0

Article URL: https://lists.nanog.org/archives/list/nanog@lists.nanog.org/thread/YDTTFIWTVGTLOUNLUXL6VNKWOIEDJ37Q/

Comments URL: https://news.ycombinator.com/item?id=47361738

Points: 1

# Comments: 0

Article URL: https://blog.rust-lang.org/2026/03/02/2025-State-Of-Rust-Survey-results/

Comments URL: https://news.ycombinator.com/item?id=47361718

Points: 1

# Comments: 0

Article URL: https://www.servethehome.com/lenovo-thinkstation-pgx-review-the-nvidia-gb10-128gb-ai-workstation-goes-corporate/

Comments URL: https://news.ycombinator.com/item?id=47361716

Points: 1

# Comments: 0

Article URL: https://phys.org/news/2026-03-sun-stellar-twins-galaxy-center.html

Comments URL: https://news.ycombinator.com/item?id=47361703

Points: 1

# Comments: 0

Article URL: https://www.youtube.com/watch?v=tc4ROCJYbm0

Comments URL: https://news.ycombinator.com/item?id=47361680

Points: 1

# Comments: 0

Article URL: https://github.blog/changelog/2026-03-12-rest-api-version-2026-03-10-is-now-available/

Comments URL: https://news.ycombinator.com/item?id=47361677

Points: 1

# Comments: 0

The flaws can be exploited to manipulate data and bypass security restrictions, potentially leading to code execution.

The post Chrome 146 Update Patches Two Exploited Zero-Days appeared first on SecurityWeek.

Article URL: https://github.com/wizwand/autoexp

Comments URL: https://news.ycombinator.com/item?id=47361666

Points: 1

# Comments: 0

Article URL: https://arstechnica.com/gadgets/2026/03/hp-has-new-incentive-to-stop-blocking-third-party-ink-in-its-printers/

Comments URL: https://news.ycombinator.com/item?id=47361645

Points: 2

# Comments: 0

Article URL: https://asciinema.org/a/834560

Comments URL: https://news.ycombinator.com/item?id=47361639

Points: 1

# Comments: 1

Hey HN,

I'm a heavy Obsidian user.

I recently got tired of the two usual sync tradeoffs:

1. File-based sync (iCloud/Dropbox/Syncthing) that leaves you waiting for changes to propagate, or hands you a "conflicted copy." 2. Self-hosted setups (like CouchDB) that need touching VMs and dockerized databases to sync markdown.

So I built YAOS: a local-first, real-time sync engine for Obsidian.

Self-hosting OSS should have better UX.

You can deploy the backend to your own Cloudflare account in one click. It fits comfortably in Cloudflare's free tier (costing $0/month for normal personal use), and requires absolutely no terminal interaction, no SSH, and no env files.

You can try it out right now: https://github.com/kavinsood/yaos

How it works under the hood:

- Text sync uses Yjs CRDTs. It syncs real-time keystrokes and cursors rather than treating the vault as a pile of files to push around later. - Each vault maps to a Cloudflare Durable Object, giving you a low-latency, single-threaded coordinator at the edge. - The backend uses a chunked Checkpoint + Delta-Journal MVCC storage engine on top of the DO's SQLite storage. - Attachments sync separately via R2 (which is optional—text sync works fine without it).

The hardest part of this project was bridging Obsidian's synchronous UI and noisy OS file-watchers with the in-memory CRDT graph. I had to build a trailing-edge snapshot drain that coalesces rapid-fire IO bursts (like running a find-and-replace-all) into atomic CRDT transactions to prevent infinite write-loops.

The current design keeps a monolithic CRDT per vault. This is great for normal personal notes, but has a hard memory ceiling (~50MB of raw text). I chose this tradeoff because I cared more about fast, boringly reliable real-time ergonomics than unbounded enterprise scale.

I also wrote up engineering notes on the tricky parts (like handling offline folder rename collisions without resurrecting dead files) on GitHub.

I've spent the last three weeks doing brutal QA passes to harden mobile reconnects, IndexedDB quota failures, and offline split-brain directory renames.

I'd love feedback on the architecture, the code, or the trade-offs I made. I'll be hanging out in the thread to answer questions!

Comments URL: https://news.ycombinator.com/item?id=47361635

Points: 2

# Comments: 0

Article URL: https://en.wikipedia.org/wiki/Shiny_object_syndrome

Comments URL: https://news.ycombinator.com/item?id=47361611

Points: 1

# Comments: 0

Article URL: https://twitter.com/tobi/status/2032212531846971413

Comments URL: https://news.ycombinator.com/item?id=47361598

Points: 2

# Comments: 0

Article URL: https://twitter.com/wordpress/status/2032291317871468554

Comments URL: https://news.ycombinator.com/item?id=47361557

Points: 1

# Comments: 0

Article URL: https://developer.nvidia.com/blog/introducing-nvfp4-for-efficient-and-accurate-low-precision-inference/

Comments URL: https://news.ycombinator.com/item?id=47361553

Points: 1

# Comments: 0