[gHashTag]

What

The FPGA course ("From zero to a chip") stated bench facts that the bench SSOT contradicts (t27 fpga/HARDWARE_SSOT.md, measured 2026-10-08: the bench board is an ALINX AX7203, xc7a200tfbg484-2). This PR corrects the two lessons and adds the X accounts to the course lead.

Where	Before	After
Lesson 1 text	"Our bench chip is an AMD Artix-7 XC7A200T."	names the board and package: ALINX AX7203, 484-ball FBG484
Lesson 1 task	"In the spec frame of our bench board, find LUT_TOTAL"	the frame is the Wukong board file (same die); its LUT_TOTAL is AMD's logic cell count, 1.6 per LUT6, not a LUT count
Lesson 3 text	"the XC7A200T on our board has 676 solder balls"	484 balls on the AX7203; the pin map draws the same die in the 676-ball FGG676 of the Wukong board
Course lead	(no X)	"On X: x.com/t27_dev and x.com/t27_lang."

The Russian bundle (i18n/course.ru.json) carries the same numbers as each English field. Generated files, lesson pages and the two course cards (EN, RU) are regenerated by the existing generators; nothing generated was edited by hand.

The drift rules that would have caught this are law in Trinity-S-AI-DNA/t27#13239 (specs/tri/course/audit.t27, PR Trinity-S-AI-DNA/t27#13241); the audited strings fail its tests verbatim and the new wording passes.

Not in this PR

• The course post. specs/policy/course_post.t27 requires a blog body in the same PR. The sanctioned route is a .t27 entry with a generated body (t27c's TypeScript generator); that generator cannot be run on this Mac (load rule), and a hand-typed TS body is forbidden by the only-t27 rule. So T27 work report is expected to stay red on post-per-pr until the post lands; it is tracked in Course 1 lessons 1 and 3: bench-board facts (AX7203, FBG484, logic cells) and the author's X handles #1804.
• Clickable X links and the mastery page: Course page: read the mastery law and link the X handles (blocked: foreign TSX) #1805 (blocked, Course.tsx is hand-written TS).

Closes #1804
Refs #1478

{
  "version": 1,
  "head_sha": "0364afe33a5210683a2b4e07a1b57fa8e42d4927",
  "summary": "Two FPGA course lessons now name the bench board (ALINX AX7203, 484-ball FBG484) instead of implying the Wukong's 676-ball package, lesson 1 says LUT_TOTAL is a logic cell count, and the course lead lists x.com/t27_dev and x.com/t27_lang.",
  "changes": [
    "apps/website/specs/course/course.t27: SAY_LEAD, LESSON_TEXTS[0], LESSON_TASKS[0], LESSON_TEXTS[2].",
    "apps/website/i18n/course.ru.json: the four matching Russian fields, with the same numbers as the English ones.",
    "Regenerated by scripts/course-from-spec.mjs: src/lib/course.generated.ts and public/learn/course.t27.",
    "Regenerated by scripts/course-pages.mjs: the EN and RU pages of the course, fpga-is-a-grid and pins-and-package, the two course cards and learn/cards.json."
  ],
  "tests": [
    {
      "command": "cd apps/website && node scripts/course-from-spec.mjs --check",
      "status": "passed",
      "result": "rc 0: 11 courses up to date, 297 lesson specs compile, spec tests 86, asserts 401 hold; the Russian bundle passes the same-numbers check.",
      "evidence": "Local run on the PR head before push."
    },
    {
      "command": "cd apps/website && node scripts/course-pages.mjs --check",
      "status": "passed",
      "result": "up to date: 616 pages for 11 courses, 616 cards, learn/sitemap.xml (after one regenerate that redrew 2 cards).",
      "evidence": "Local run on the PR head before push."
    },
    {
      "command": "lefthook pre-commit and pre-push (explorer-bilingual-contract, queen-bilingual-contract, course-ladder-contract, translations-contract, english-only)",
      "status": "passed",
      "result": "all green on commit and push.",
      "evidence": "Local hook output on commit 0364afe33a."
    },
    {
      "command": "npm run typecheck && npm run test:course && npm run check:course-pages (the node --test half)",
      "status": "not_run",
      "result": "Not run on this Mac: tests and builds run in CI only (load rule). CI runs them on this PR.",
      "evidence": "CI checks on this PR."
    }
  ],
  "limitations": [
    "The course post required by specs/policy/course_post.t27 is not in this PR, so the post-per-pr rule is expected to fail until it lands (see Not in this PR).",
    "The X handles are plain text in the lead, not links: the course page has no link slot that a spec can fill (#1805).",
    "Only the three drifts found in the audit at ca7ec465e5 are fixed; no other lesson was re-checked against the bench beyond the drift rules in Trinity-S-AI-DNA/t27#13239."
  ],
  "tags": ["Course", "Engineering", "Verification"],
  "blog": {
    "title": "The course said 676 balls; the bench board has 484",
    "summary": "An audit against the bench SSOT found three sentences in the FPGA course that were true of a different board, and the rule that catches them is now a t27 spec.",
    "outline": [
      "Lessons 1 and 3 described the XC7A200T on the Wukong board (FGG676, 676 balls) as if it were the bench chip, which sits on an ALINX AX7203 in the 484-ball FBG484, and called a logic cell count a LUT count.",
      "The fix names the board and package in each sentence and keeps the Wukong frame and pin map, saying which board they draw; a drift rule (trigger, required context) in specs/tri/course/audit.t27 fails the old strings and passes the new ones, and its package trigger is ' 676' with a space so it does not fire inside 1.6467602581.",
      "The audit has no tri glue yet, so it runs only through its own tests; the clickable X links and a mastery page wait on a course page that a spec can drive."
    ]
  }
}

🤖 Generated with Claude Code