Skip to content

Releases

Performance fix verification, 2026-10-18

Choose a tag to compare

@luke2x luke2x released this 18 Oct 19:55

Performance helper evidence (2026-10-18)

These are synthetic CPU benchmarks of the production helpers, not an application FPS, native Hermes, or heap benchmark. The script extracts the production helper source from the supplied Git refs and strips TypeScript. It does not execute caller components or parsing libraries outside the isolated helper.

Environment: {'node': 'v26.10.0', 'platform': 'darwin', 'arch': 'arm64', 'cpu': 'Apple M5'}

Method: five warmup rounds; 30 alternating before/after samples. GC runs outside timed sections when available. Input objects/strings are reused. Each large-preview sample batches 20 calls; short-preview samples batch 2,000. Output equivalence is asserted before timing. Raw per-call timings are in results.json.

Run from a repository checkout containing the refs (fetch the corresponding luke2x/t3code branches first):

node --expose-gc performance-bench.mjs d9eb2b6 98cc2b4 374b83c 59090b8

tree: median 3.652625 ms before; 0.333959 ms after.
preview: median 0.990725 ms before; 0.003665 ms after.
previewShort: median 0.000103 ms before; 0.000123 ms after.
html: median 64.355666 ms before; 2.788375 ms after.

Workloads:

  • Tree: 30,000 files under 21 expanded ancestors; measures flattening only, not tree construction or native rendering. A separate 200,000-file directory case throws RangeError before and returns 200,001 rows afterward. That wide case is a stress test, not a claimed field report.
  • Preview: a 262,144-character completed result, normalized for the existing 280-character display limit; plus a short, common-case progress message.
  • HTML: 3,000 inline code ranges with 6,000 HTML image tags (literal tags inside code and converted tags outside). Measures markdownHtmlContent, not the full markdown parser or layout.

The independent mobile visibility PR has React/jsdom behavior tests for text-only rows, sharing one observer across multiple videos, cleanup, and explicit visibility outside a feed. Its LegendList viewability boundary is mocked. No native scrolling recording, native frame-time measurement, or Hermes timing was collected because this host has no connected device, Android SDK/emulator, or full Xcode simulator tooling. It is a subscription-lifetime fix; these helper benchmarks are not evidence of its frame-rate impact.

PR evidence: Windows PowerShell environment timeout

Choose a tag to compare

Evidence for a pingdotgg/t3code PR about the Windows startup PowerShell timeout (pingdotgg#4403). Not a release.

PR evidence: streamed browser tab frame rate

Choose a tag to compare

Evidence for a pingdotgg/t3code PR about the streamed browser tab frame rate (pingdotgg#17090). Not a release.

PR evidence: ACP terminal output cost

Pre-release

Choose a tag to compare

Evidence for a pingdotgg/t3code PR about ACP terminal output (pingdotgg#19052). Not a release.

PR evidence: context meter million boundary

Choose a tag to compare

@luke2x luke2x released this 09 Oct 16:40
534f999

Screenshots for pingdotgg#17539.

Focused bug fix verification, 23 October 2026

Choose a tag to compare

@luke2x luke2x released this 23 Oct 16:32
61b9790

Synthetic regression evidence for Claude session identity, bounded human-turn history, and Azure DevOps PAT discovery. Verification limits and provenance are documented in README.md.

PR 24168 review follow-up evidence

Choose a tag to compare

@luke2x luke2x released this 24 Oct 23:43
61b9790

PR pingdotgg#24168 review follow-up

Baseline: 91b52c8 (Julius’s latest commit). Follow-up: a10d7c5.

  • delivery-before: current late-subscriber regression against the baseline ProviderSessionManager.ts. The current manager file was restored in a finally block. Expected failure: no adopted-session update persisted.
  • delivery-after: all 51 manager tests pass on the fixed source, including eight delivery/release/rewind cases. Queue/Deferred/persisted-event synchronization; no added sleeps or polling.
  • attempt-fork-before/after: initial regression run before the attempt-ID and synthetic fork-fixture corrections, then all 50 tests in the two files pass. These logs predate the additional manager persistence coordination.
  • ingestor-after: 24 tests pass, including preserving rewind side effects when a stale snapshot is skipped.
  • check-changed: final cumulative PR checks pass: formatting, lint, both knip checks, generated GraphQL, 21 affected-package typechecks and 457 focused tests in 10 files. Existing unused-variable warnings remain.

Environment: macOS Apple Silicon, Node 26.10.0. Synthetic runtime fixtures and in-memory database tests. No fresh authenticated Claude /clear or native-client recording was performed. The fork fixture explicitly identifies its synthetic adaptation. Paths are redacted; no personal database, credentials, conversation content, or environment settings are included.

Made with GPT-6.1 Sol through the Codex harness in T3 Code.

PR evidence: macOS test capture error message

Choose a tag to compare

@luke2x luke2x released this 12 Oct 23:51
61b9790

Screenshots for a pingdotgg/t3code pull request. Not a release of T3 Code.

PR evidence: limit recovery after idle wakes

Choose a tag to compare

@luke2x luke2x released this 11 Oct 21:41
61b9790

Screenshots and recordings for pingdotgg#17862. Not a software release.

LaTeX rendering prototype screenshots

Pre-release

Choose a tag to compare

@luke2x luke2x released this 08 Oct 19:19
61b9790

Before/after evidence for a local LaTeX rendering prototype in T3 Code's built-in Browser panel.

Each pair uses the same assistant Markdown, viewport, theme, and baseline commit (e6113f2606c0717feab142d906c1e995afaabe6e). The after captures add the uncommitted local math prototype. This does not represent a published LaTeX PR or a finished opt-in feature.

The source repeats the three matrix questions from the reported rendering problem, without private conversation details. These are unedited native captures.

  • before.png / after.png: original matrix example, 900 × 900.
  • showcase-before.png / showcase-after.png: integrals, series, limits, fractions, roots, Greek letters, aligned working, and a piecewise function, 900 × 1150.

The expanded app example renders ten expressions (six display and four inline) with fonts loaded and no KaTeX errors. Focused parser, chat-renderer, and incremental-rendering checks pass: 114 tests across three files.