frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Open in hackernews

Show HN: VectorJSON – O(n) streaming parser to handle LLM JSON outputs

https://github.com/teamchong/vectorjson
1•teamchong•1h ago

Comments

teamchong•1h ago
I built this after hitting GC stalls parsing streaming tool calls in my AI agent. LLM outputs are getting large — code edits, file writes, 50-200KB JSON payloads.

Every AI SDK I looked at (Vercel, Anthropic, TanStack, OpenClaw) does `buffer += chunk; JSON.parse(buffer)` on every token. That's O(n²) — a 100KB tool call arrives in ~8000 chunks, and each chunk re-parses the entire accumulated buffer from scratch. The cumulative parse time adds up to 13.4 seconds for the Anthropic SDK. Each intermediate buffer string and parsed object becomes garbage immediately, thousands of short-lived allocations that put constant pressure on the GC.

VectorJSON scans only the new bytes on each chunk. O(n) total — same payload, 6.6ms. Parsing happens in WASM linear memory, so no JS objects are created until you access a field.

Built on zimdjson (Zig port of simdjson by @travisstaloch) compiled to WASM. Fields materialize lazily through a Proxy — if you only read 3 fields from a 100KB payload, the other 97% never touches the JS heap. Deep comparison runs entirely in WASM memory — 2-4× faster than recursive JS deepEqual with 24× less heap pressure.

Not a replacement for JSON.parse — for single-shot full materialization, JSON.parse is faster (it's optimized C++ in V8). VectorJSON is built for streaming, partial access, and deep comparison.

Ladybird: Closing this as we are no longer pursuing Swift adoption

https://github.com/LadybirdBrowser/ladybird/issues/933
1•thewavelength•54s ago•0 comments

Ukraine's no man's land is the future of war

https://www.ft.com/content/9c7542d4-1b2b-4062-b53d-bb02b5f12851
2•danielmorozoff•1m ago•0 comments

Stress-hardening behaviour of biofilm streamers

https://www.nature.com/articles/s41467-025-64557-x
1•PaulHoule•4m ago•0 comments

Chinese New Year v.s. Lunar New Year

1•dogclaw•4m ago•0 comments

Smart model routing for agentic coding

https://knowmatic-lab.xyz/
1•tmickleydoyle•4m ago•0 comments

PlanetScale vs. Supabase Benchmarks

https://planetscale.com/benchmarks/supabase
1•aarvinroshin•5m ago•0 comments

An Exercise in Agentic Coding: AV1 Encoder from Scratch in Rust

https://caricio.com/blog/an-exercise-in-agentic-coding-av1-encoder-from-scratch-in-rust/
1•rjcaricio•5m ago•0 comments

Show HN: Faultline – Open-source AI agent for infrastructure debugging

https://github.com/chatwoot/faultline
1•pranav_rajs•6m ago•1 comments

Missing skiers found dead following deadliest U.S. avalanche in 45 years

https://abcnews.com/US/backcountry-skiers-missing-avalanche-northern-california-authorities/story...
2•sizzle•10m ago•0 comments

Harrier: A high-performance Rust hashmap using SIMD-accelerated cuckoo hashing

https://github.com/kushalthaman/harrier
1•kushalt•10m ago•1 comments

Show HN: Generate baseline Kubernetes NetworkPolicies from rendered manifests

https://github.com/dormstern/segspec
1•dormstern•10m ago•0 comments

OpenAI's acquisition of OpenClaw signals the end of the ChatGPT era

https://venturebeat.com/technology/openais-acquisition-of-openclaw-signals-the-beginning-of-the-e...
2•gmays•11m ago•0 comments

Ask HN: In Cursor/agents, do plugins hide MCP tools from the main agent?

1•azebazenestor•13m ago•0 comments

Show HN: Designing package namespacing for a new language (Coi)

1•io_eric•15m ago•0 comments

Oral History of Michael J. Flynn [video]

https://www.youtube.com/watch?v=OD2uE9X9BPs
1•matt_d•15m ago•0 comments

Show HN: Build Minecraft mods and servers in the browser

https://www.orcaengine.ai/
1•ekduman•16m ago•0 comments

Micron's PCIe 6.0 SSD Hits Mass Production at 28 GB/S

https://www.tomshardware.com/pc-components/ssds/worlds-first-pcie-6-0-ssd-enters-mass-production-...
2•m463•19m ago•0 comments

Show HN: HiddenState – How I keep up with 500+ ML papers a day

https://hiddenstate.io/archive/2026-02-18
1•CosmoSantoni•19m ago•1 comments

Britain Lost a Quarter of all pubs (14,000 Pubs) in 13 Years

https://laurenleek.substack.com/p/britain-lost-14000-third-places-they
2•m463•20m ago•1 comments

All Look Same?

https://alllooksame.com/
4•mirawelner•21m ago•0 comments

Martial arts robots dazzle at 2026 Spring Festival Gala [video]

https://www.youtube.com/watch?v=mUmlv814aJo
3•lisper•21m ago•0 comments

Show HN: Geneclaw – An AI agent framework that safely evolves its own code

https://github.com/Clawland-AI/Geneclaw
1•geneclawai•22m ago•1 comments

GitSyncMarks – Browser extension that syncs bookmarks to your own GitHub repo

https://github.com/d0dg3r/GitSyncMarks
2•d0dg3r•27m ago•1 comments

Owning Your Data

https://www.coryd.dev/posts/2026/owning-your-data
3•cdrnsf•28m ago•0 comments

Andrew Ranken, Whose Drumming Powered the Pogues, Dies at 72

https://www.nytimes.com/2026/02/11/arts/music/andrew-ranken-dead-the-pogues.html
2•bookofjoe•28m ago•1 comments

Astrolabe

https://en.wikipedia.org/wiki/Astrolabe
2•jhncls•28m ago•0 comments

The A.I. Disruption We've Been Waiting for Has Arrived

https://www.nytimes.com/2026/02/18/opinion/ai-software.html
4•gyomu•30m ago•0 comments

Productivity App to auto categorize your work and improve your workflow

https://dreamdimension.net/deepfocusapp/
2•dreamdimension•30m ago•1 comments

I was banned from the Wikipediocracy forum after unmasking a pro-CCP doxxer

https://xcancel.com/Liltjay08Foo/status/2023735372136464471
3•kurtreed2•31m ago•1 comments

Persona: Controlling LLM Personality with Vector Algebra

https://arxiv.org/abs/2602.15669
2•mldev_exe•31m ago•0 comments