Skip to content

term: eight FPGA bench recordings - x7-audit, x7-mutate, fpga-status, x7-refs, fpga-pins, fpga-batch, fpga-tmpcheck, fpga-tools - #1329

Merged
gHashTag merged 1 commit into
mainfrom
feat/term-x7-audit-mutate
Oct 4, 2026
Merged

gHashTag merged 1 commit into
mainfrom
feat/term-x7-audit-mutate

Conversation

@gHashTag

@gHashTag gHashTag commented Oct 4, 2026

Copy link
Copy Markdown
Owner

Eight new recording pages under apps/website/public/term/, made with tri cast publish (scrub, check, card, X-sized GIF), for the FPGA-engineer series on @t27_dev:

id command(s) what it shows
tri-x7-audit tri x7-audit CRC 12/12 over 6 references, ECC 32520/32520 frames, FAR walk = Vivado FDRI, pins 82/82, segbits 4224/4224, writer 16/16, fasm 22/22, gold 5/5: audit: CLEAN
tri-x7-mutate tri x7-mutate packets.t27 5 planted bugs (CRC poly bit, CRC address width, LFRM NOOP count, COR0 bit 0, type-2 count): all 5 caught by tests and sweeps
tri-fpga-status tri fpga-status, tri fpga-usb UART direct (no hub), Digilent JTAG, lock free, three x7-board runs PASS 51840/51840
tri-x7-refs tri x7-part, tri x7-refs, tri x7-bench 5 part table = prjxray-db part.json, 6 references present, bitwalk 84.4 ms best of 5
tri-fpga-pins tri fpga-pins fpga/portable/trinet_node_core.v 5 specs pin the RTL core by sha256, 5 generators read them: radius 5 specs, 1 level
tri-fpga-batch tri fpga-batch-rehearsal 26 checks off-board: 11 loss/fault cases recover, lie/duplicate/wrong key stop the run, PASS
tri-fpga-tmpcheck tri fpga-tmpcheck 112 /tmp files pinned by specs: all gone after boot, all kept, 0 lost
tri-fpga-tools tri fpga-tools 14/14 bench tool self-tests pass

Every cast: exit codes all 0, home directory scrubbed to ~, 0 key hits. Gallery public/term/index.html rebuilt (30 recordings).

The eleven trios cards these recordings run carry CAST, generated by scripts/tools-from-trios-tri.mjs at the pinned BrowserOS 7366096248df, not edited by hand. fpga-usb now points at tri-fpga-status (which also runs it) instead of tri-fpga-wire; that is the generator's pick.

{
  "version": 1,
  "head_sha": "c3f69f76462ea41d04dcd1c50f4175f64e1da7a2",
  "summary": "Eight more FPGA tri commands are recorded for the FPGA-engineer series: the whole bitstream audit, mutation testing of the packet spec, the bench status, the reference bitstreams, the blast radius of an RTL edit, the batched wire protocol rehearsal, pinned /tmp inputs and the bench tools' self-tests; eleven trios tool cards now name their recording.",
  "changes": [
    "apps/website/public/term/{tri-x7-audit,tri-x7-mutate,tri-fpga-status,tri-x7-refs,tri-fpga-pins,tri-fpga-batch,tri-fpga-tmpcheck,tri-fpga-tools}/: eight new recording pages (index.html, session.cast, session.gif, card.png, meta.json) written by tri cast publish, and the gallery index rebuilt to 30 recordings",
    "apps/website/public/t27/files/specs/tools/trios/tri/{x7-audit,x7-mutate,x7-part,x7-refs,x7-bench,fpga-status,fpga-usb,fpga-pins,fpga-batch-rehearsal,fpga-tmpcheck,fpga-tools}.t27: CAST added or moved by scripts/tools-from-trios-tri.mjs at the pinned BrowserOS 7366096248df, no hand edits",
    "apps/website/public/tools/spec-tools.json, spec-tools-trios.json and the agents/crons/functions/skills catalogs: regenerated by agents-from-specs.mjs, 35 cards with a recorded run",
    "apps/website/scripts/agents-from-specs.test.mjs: withCast pinned at 35 instead of 25, with the 27 trios cards from 22 recordings explained"
  ],
  "tests": [
    {
      "command": "npm run check:tools-coverage",
      "status": "passed",
      "result": "every tri command has a card, every card a command, every unambiguous recorded command its CAST; 596 commands, 35 with a recording",
      "evidence": "local run of this head, 2026-10-04"
    },
    {
      "command": "npm run check:tools",
      "status": "passed",
      "result": "607 cards typecheck",
      "evidence": "local run of this head, 2026-10-04"
    },
    {
      "command": "node --test scripts/agents-from-specs.test.mjs",
      "status": "passed",
      "result": "53 of 53 pass",
      "evidence": "local run of this head, 2026-10-04"
    },
    {
      "command": "npm run check:tools-site && npm run check:agents && npm run check:onboarding && npm run check:subpath-urls && npm run check:shared-core && npm run test:tri-commands",
      "status": "passed",
      "result": "all six exit 0",
      "evidence": "local run of this head, 2026-10-04"
    },
    {
      "command": "tri cast check on the eight new session.cast files",
      "status": "passed",
      "result": "8 of 8: every exit code 0, key hits 0, home directory shown as ~",
      "evidence": "local run before tri cast publish, 2026-10-04"
    },
    {
      "command": "npx vite build (apps/website)",
      "status": "not_run",
      "result": "not run locally; the website CI build covers it",
      "evidence": "no local build was run for this head"
    }
  ],
  "limitations": [
    "tri fpga-status reports the board as NOT LOADED: its last load record is an earlier run that ended with exit -6; this recording drives no board.",
    "tri fpga-tmpcheck finds all 112 pinned /tmp files gone (the directory is emptied at boot) and every one kept in artifacts/model_inputs/blobs; it does not restore them in this recording.",
    "tri x7-audit and tri x7-mutate were recorded with X7_T27C pointing at a t27c built from t27 master, which carries t27#5594; the default assertfix build had been removed by the cache cleaner.",
    "Nothing was posted from this PR. The X posts are scheduled by hand after each page answers 200 on t27.ai."
  ],
  "tags": [
    "Website",
    "FPGA",
    "Tooling"
  ],
  "blog": {
    "title": "Eight more FPGA commands, recorded: audit, mutants, bench status, pins",
    "summary": "Eight tri commands an FPGA engineer runs around a 7-series bitstream path are now recorded with real output, from a full audit to planting bugs the tests must catch.",
    "outline": [
      "tri x7-audit re-runs every bitstream sweep against 6 reference bitstreams and ends audit: CLEAN, and tri x7-mutate plants 5 bugs in the packet spec and catches all 5.",
      "tri fpga-status and tri fpga-usb read the bench in one screen, tri fpga-pins names every spec an RTL edit will touch before the edit, and tri fpga-tmpcheck keeps 112 pinned /tmp inputs.",
      "tri fpga-batch-rehearsal runs 26 off-board checks of the batched wire protocol and tri fpga-tools runs the self-tests of 14 bench tools, 14 of 14 passing."
    ],
    "cast": "term/tri-x7-audit/session.cast",
    "reproduce": [
      "tri x7-audit",
      "tri x7-mutate packets.t27",
      "tri fpga-pins fpga/portable/trinet_node_core.v"
    ]
  }
}

🤖 Generated with Claude Code

… x7-refs, fpga-pins, fpga-batch, fpga-tmpcheck, fpga-tools

Eight new share pages under public/term/, each a recorded run with real output,
all exit codes 0, scrubbed and checked by tri cast check:

- tri-x7-audit: every bitstream sweep re-run, audit: CLEAN
- tri-x7-mutate: 5 planted bugs in packets.t27, all 5 caught
- tri-fpga-status: tri fpga-status + tri fpga-usb, UART direct, no hub
- tri-x7-refs: tri x7-part + x7-refs + x7-bench, 6 references, 84.4 ms
- tri-fpga-pins: blast radius of trinet_node_core.v, 5 specs, 1 level
- tri-fpga-batch: tri fpga-batch-rehearsal, 26 checks, PASS
- tri-fpga-tmpcheck: 112 pinned /tmp files, 0 lost
- tri-fpga-tools: 14/14 tool self-tests pass

The trios tri cards for the eleven recorded commands carry CAST (generated
by tools-from-trios-tri.mjs at BrowserOS 7366096248df); fpga-usb now points
at tri-fpga-status, which also runs it. spec-tools*.json regenerated:
35 tools with a recorded run (27 trios + 8 t27); the test count follows.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@github-actions github-actions Bot added the status:in-progress 🔵 Agent working label Oct 4, 2026
@gHashTag
gHashTag merged commit e62e6ae into main Oct 4, 2026
32 of 37 checks passed
@github-actions github-actions Bot added status:completed Done and removed status:in-progress 🔵 Agent working labels Oct 4, 2026
github-actions Bot added a commit that referenced this pull request Oct 4, 2026
term: eight FPGA bench recordings - x7-audit, x7-mutate, fpga-status, x7-refs, fpga-pins, fpga-batch, fpga-tmpcheck, fpga-tools (#1329)

Eight new share pages under public/term/, each a recorded run with real output,
all exit codes 0, scrubbed and checked by tri cast check:

- tri-x7-audit: every bitstream sweep re-run, audit: CLEAN
- tri-x7-mutate: 5 planted bugs in packets.t27, all 5 caught
- tri-fpga-status: tri fpga-status + tri fpga-usb, UART direct, no hub
- tri-x7-refs: tri x7-part + x7-refs + x7-bench, 6 references, 84.4 ms
- tri-fpga-pins: blast radius of trinet_node_core.v, 5 specs, 1 level
- tri-fpga-batch: tri fpga-batch-rehearsal, 26 checks, PASS
- tri-fpga-tmpcheck: 112 pinned /tmp files, 0 lost
- tri-fpga-tools: 14/14 tool self-tests pass

The trios tri cards for the eleven recorded commands carry CAST (generated
by tools-from-trios-tri.mjs at BrowserOS 7366096248df); fpga-usb now points
at tri-fpga-status, which also runs it. spec-tools*.json regenerated:
35 tools with a recorded run (27 trios + 8 t27); the test count follows.

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
gHashTag added a commit that referenced this pull request Oct 4, 2026
main e62e6ae (#1329) rewrote spec-agents/crons/functions/skills/tools.json
for new FPGA tool cards; this branch had added the providers rung to their
ladder. All five are generated files, so the conflict was resolved by taking
main's copy and re-running scripts/agents-from-specs.mjs: each now carries
main's content plus ladder.providers = 7. agents-from-specs --check passes.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
gHashTag added a commit that referenced this pull request Oct 4, 2026
main moved by 20+ commits since 8a8f0d7 (providers tab #1302, brain health
#1328, FPGA recordings #1329, blog #1309). Clean merge. check:queen-runners,
queen-spec-sync, queen-contrast and queen-languages pass on the merged tree;
check:queen-honesty fails 3 on inputs byte-identical to main (Queen.tsx,
QueenCombBabylon.tsx), so that is main's state, not this branch.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant