Updated Jul 4

• ♦ 8065689

  perf(gates): add per-frame allocation gate (G3) + codify the gate suite

• ♦ 8d055ea

  Merge pull request #36 from danReynolds/danreynolds/perf-alloc-gate

Builds G3 (the last deferred perf gate) and closes the codification gap the assessment surfaced: the gates existed but nothing told anyone — human …

Updated Jul 4

• ♦ 9c69632

  iOS: ship the native comrak bridge through native assets (#23)

Ships flark's native comrak bridge on iOS through the same native-assets
build hook as macOS/Android/Linux — no CocoaPods podspec, no prebuilt

Builds G3 (the last deferred perf gate) and closes the codification gap the assessment surfaced: the gates existed but nothing told anyone — human …

Updated Jul 4

Interactive dependency upgrader for npm, yarn, pnpm & bun — changelog viewer, vulnerability audit, monorepo-ready, zero-config.

TypeScript 24 Updated Jul 4

🚀🤖 Crawl4AI: Open-source LLM Friendly Web Crawler & Scraper. Don't be shy, join here: discord.gg/jP8KfhDhyN

Python 70.9k Updated Jul 3

Updated Jul 4

• ♦ 23f9020

  Exp 215: reject persistent executeWrite result-buffer slot (#230)

While writing up exp 211 — which accepted persistent per-isolate scratch slots for the reader-side queryBytes() after exp 108 had reject…

While writing up exp 211 — which accepted persistent per-isolate scratch slots for the reader-side queryBytes() after exp 108 had reject…

Updated Jul 4

Updated Jul 4

Updated Jul 4

• ♦ b1c68c6

  Exp 214: reject write result direct-read prototype

executeWrite() decodes a 16-byte native resqlite_write_result on every public write by first wrapping the pointer in a typed list and th…

• ♦ fb56668

  Exp 214: address review feedback

executeWrite() decodes a 16-byte native resqlite_write_result on every public write by first wrapping the pointer in a typed list and th…

Updated Jul 4

• ♦ c8389c5

  Address code review: scope goldens, fix stale iOS text, harden checks

Ships flark's native comrak bridge on iOS through the same native-assets
build hook as macOS/Android/Linux — no CocoaPods podspec, no prebuilt

Updated Jul 4

Updated Jul 4

• ♦ 87ce466

  perf(framework): drop redundant O(n) child-order copy in reconcile

• ♦ 26ee004

  Merge pull request #35 from danReynolds/danreynolds/reconcile-redunda…

Closes out the §2 retained-tree-tax investigation of the perf pass. Verdict: confirmatory — the retained-tree machinery is lean, and this lands the…

Closes out the §2 retained-tree-tax investigation of the perf pass. Verdict: confirmatory — the retained-tree machinery is lean, and this lands the…

Updated Jul 4

• ♦ d9bb4bf

  perf(web): first-load bundle-size gate for the served-browser client …

• ♦ ab1e6af

  Merge pull request #34 from danReynolds/danreynolds/bundle-size-gate

G6 — first-load bundle-size gate. web-capture validated the dart2js client exists, not its size, so a stray import could silently add 100s of KB to…

G6 — first-load bundle-size gate. web-capture validated the dart2js client exists, not its size, so a stray import could silently add 100s of KB to…

Updated Jul 3

• ♦ a99beac

  perf(image): image byte category + inline-image encoder bench (G2)

• ♦ 096aca9

  Merge pull request #33 from danReynolds/danreynolds/image-bench

G2 (terminal side) — image-pipeline coverage. Images are the largest per-frame payload, but they were invisible to every byte-level analysis.

G2 (terminal side) — image-pipeline coverage. Images are the largest per-frame payload, but they were invisible to every byte-level analysis.

Updated Jul 3

• ♦ af6ab31

  perf(serve): guard the semantics-wire anti-cliff invariant (Tier 1 ou…

• ♦ 0454325

  Merge pull request #32 from danReynolds/danreynolds/serve-semantics-gate

Tier 1 outcome — the premise didn't survive scrutiny, so this guards the invariant instead of forcing a low-value optimization.

Tier 1 was "reduce …

Tier 1 outcome — the premise didn't survive scrutiny, so this guards the invariant instead of forcing a low-value optimization.

Tier 1 was "reduce …

Updated Jul 3

• ♦ 6238c81

  perf(serve): live serve-wire profiler + gate (G1)

• ♦ 1bff63c

  review: harden the live serve-wire profiler + gate

• 1 more commit »

G1 from the perf-pass findings — the biggest coverage blind spot. The fleury serve wire was measured only synthetically (serve_wire_profile computes …