docs(conformance): section 4 of FORMATS_ON_BOARD.md points to the two TNF board runs

[dmitrii-f-t27]

Closes #903

Section 4 ("Never on this board") listed TNF without qualification. Sections 8 and 9 of the same file record two TNF board runs on 2026-09-27:

• board_runs/tnf16_board_ax7203.log: TNF16 add and mul, TNFFormat(4, 11) v2-spec, 1,035,886/1,035,886 bit-exact, fails 0, lost 0, exit 0.
• board_runs/mxdot4_board_ax7203.log: MXFP4/TNF4 block dot product, 251,616/251,616 bit-exact.

The change adds one paragraph before the TNF and BNF bullet. It says both runs are arithmetic exactness only (no clock was measured, no advantage follows), and that the bullet still holds for BNF, the other three TNF16 definitions and the LUT/Fmax figures it cites. No log, number or other section changes; the added text is ASCII.

Dmitrii Fedorov (@dmitrii-f-t27)

[tri-claw]

Summary by t27-bees

This PR updates section 4 of FORMATS_ON_BOARD.md to acknowledge two TNF board runs added in sections 8 and 9 (2026-09-27) while clarifying they only demonstrate arithmetic exactness without clock measurements. The original claim that TNF/BNF were "never on this board" is narrowed to still apply to BNF, the other three TNF16 definitions, and the cited LUT/Fmax figures.

Worth a look

• conformance/FORMATS_ON_BOARD.md:87-93: The added paragraph states "no clock was measured and no advantage follows" for the two new runs. If sections 8 or 9 later add timing data, this paragraph would become inaccurate without a corresponding update here.

---

_{Read: 1 files, +8 -0, the whole diff. Head 2fd004f. Model: openai-compatible/nvidia/nemotron-3-ultra-550b-a55b. Commands: @t27-bees review | summary | help | pause | resume. Free for every repository.}

[github-actions]

X Brain Health Check

Score: 100.0/100
Status: 🟢 HEALTHY
Threshold: 80/100

X Brain is above merge threshold

[github-actions]

🧪 Brain Stress Test Results

Score: not measured (no stress test prints a Score: line)
Status: 🟢 PASS
Threshold: 270/300

X Stress test passed

[dmitrii-f-t27]

The three red checks come from main, not from this one-paragraph change to conformance/FORMATS_ON_BOARD.md:

• doc-refs (job 114027209793): the 9 new dangling references are in AGENTS.md, CLAUDE.md and research/*.md (agents.json, gft16_ref.py, ...); none is in FORMATS_ON_BOARD.md.
• orphan-artefacts (job 114027162866): research/arxiv_gft/competition.json.
• yosys-synth (job 114027210128): the container exits with 125 before synthesis.
  main's head ea895b5 shows doc-refs and orphan-artefacts red too. This PR stays open until main is green; it does not merge on a red check.

[dmitrii-f-t27]

Update on the three red checks at the same head e592f146.

yosys-synth was not a real failure: the job died on a Docker Hub pull limit,
toomanyrequests fetching regymm/openxc7:latest, exit 125 before synthesis started.
I re-ran that job alone and it passes on the unchanged head. Checks at e592f146
are now 47 pass, 2 fail.

The two that remain are doc-refs and orphan-artefacts, and they reproduce on
main's head ea895b50 ("Document references" and "Orphaned artefacts" both failing
there), so they are inherited and not reachable from a one-paragraph edit to
conformance/FORMATS_ON_BOARD.md. doc-refs prints its own remedy for whoever owns
that fix: 9 exclusions need --update-lists, and 2 stale baseline entries need
--prune-baseline on tools/doc_refs_baseline.txt.

Position unchanged: this does not merge on a red check. It needs main green, or a
baseline re-bless in a separate PR.

Dmitrii Fedorov (@dmitrii-f-t27)

[dmitrii-f-t27]

Status at head c2986cfb: three red checks, none reachable from this one-paragraph edit to conformance/FORMATS_ON_BOARD.md.

• doc-refs (job 114088502992): the 7 dangling references in research/*.md, the 9 cross-repo exclusions missing from the lists and the 2 stale baseline entries are the same as on main (run 37817628394 on ea895b50; main's head 09662bbc has the same tree, a3140955). Two more appeared since then: AGENTS.md and CLAUDE.md name agents.json, which resolved in gHashTag/t27 until t27b vacuous: delete 22 empty generated spec stubs (Closes #8052) t27#8054 deleted .trinity/seals/agents.json (f06b12875, 2026-10-09). This PR touches neither file.
• orphan-artefacts (job 114088502875): research/arxiv_gft/competition.json and gft_competition.pdf, same as main (run 37817628378).
• Validate VIBEE Codegen (run 38010300278, both attempts): DOES NOT COMPILE: specs/tri/phi_utils_multi.tri, the moving-sample defect in codegen-corpus: compile sample moves with the run, so one commit is red and green (#851 W15) #885; the fix is fix(codegen): sample the corpus compile check from the baseline #886.

#872 fixes the first two. At its head 7cf44298, doc-refs is red only on the two agents.json references above.

Position unchanged: no merge on a red check. This merges once main is green (#872, #886) and the checks re-run green on this head.

Dmitrii Fedorov (@dmitrii-f-t27)