Skip to content
DenseChip
Learn
PracticeCode labInterview

DenseChip

Navigate

Menu

LearnPracticeCode labInterview

Tools

All toolsSV editorCode repoResume
DenseChip

Practice, courses, and chip-design tools in one focused workspace.

LearnPracticeCode labInterviewToolsLearning paths

© 2026 DenseChip

AboutContactPrivacyTerms
Skip to lesson

Course contents

  • Getting Started with AI for VLSI Engineers5
  • Just Enough AI: Models and Context15
  • Prompting13
  • Trust and Verification14
  • Using AI for Engineering Work17
  • Engineering Walkthroughs7
    • Specification to Requirements
    • RTL Review
    • Verification Plan and Assertions
    • Scoreboard Debug
    • Timing Report
    • Physical Design Reports
    • Engineering Writing
  • Practical Foundations for AI Workflows4
  • Engineering Data5
  • Retrieval-Augmented Generation6
  • Agents and Tool Use18
  • Agent Design8
  • MCP and Skills6
  • Multi-Agent Design6
  • Practical Capstone6
  • Tokens, Cost and Latency12
  • Agentic Workflows for Chip Teams18
  • Interview and Judgment10
  • AI Skills: Full Course Index
  • Working with AI Glossary
  • Working with AI Cheatsheet
Loading lesson

‹Engineering Walkthroughs
  1. Topics
  2. AI Skills for VLSI Engineers
  3. Engineering Walkthroughs
  4. Engineering Writing

Role-based VLSI Walkthroughs · Beginner

Engineering Writing

Draft a concise review handoff that preserves evidence, uncertainty, ownership, and verification status.

On this page

    Problem and evidence

    You must hand DCPeripheral findings to another engineer who has not read the chat. An assistant can turn working notes into a clear issue or review summary, but polished language can erase uncertainty. The task is to make every consequential statement traceable to evidence and every proposed action distinguishable from completed work. Begin with the decision the reader needs to make, then give the observations needed to assess it.

    plaintext
    1
    2
    3
    4
    5
    6
    7
    8
    9
    SYNTHETIC working notes W-A:
    Spec A: synchronous active-low reset; config is 8 bits; FIFO depth 4.
    RTL fragment: negedge rst_n event control; occupancy declared [1:0].
    SB-A: expected A5 / observed 00 after a sampled reset; reset branch absent
          in supplied scoreboard excerpt.
    T-B: setup arrival 1.97 ns, required 1.85 ns, slack -0.12 ns.
    P-A/P-B: overflow 20/45 track units; seed and density both changed.
    Status: inspection and hand calculations only; no checks executed.
    Open: response timing and reset cancellation of pending requests.

    Give the assistant a writing contract

    plaintext
    1
    2
    3
    4
    5
    6
    PROMPT:
    Draft a review note from W-A for the DCPeripheral team.
    Lead with the requested decision. Preserve artifact IDs and units.
    Separate observations, hypotheses, proposed changes, and future checks.
    Use role owners, not invented people or addresses.
    Do not claim a test passed, a root cause was proven, or a message was sent.

    Before reading the example response, notice how different the evidence strengths are. The reset event control and occupancy width directly conflict with the supplied contract. The scoreboard diagnosis depends on whether the excerpt contains the whole relevant update path. The timing violation is arithmetic under a stated setup context. The physical comparison is an observation with unresolved causes. Writing all four as confirmed root causes would mislead the next engineer even if the prose sounded confident.

    Worked handoff draft

    plaintext
    1
    2
    3
    4
    5
    6
    7
    8
    9
    10
    11
    12
    13
    14
    15
    16
    17
    18
    19
    20
    21
    22
    SYNTHETIC assistant response, edited teaching draft; not sent:
    Decision requested: review the reset/count correction proposals and resolve
    the pending-response reset contract before extending the DV plan.
    
    Observed: Spec A requires rising-edge synchronous reset. The supplied RTL
    uses negedge rst_n and a two-bit occupancy count that cannot represent four.
    T-B has setup slack 1.85 - 1.97 = -0.12 ns at synthetic_slow.
    
    Hypotheses: SB-A's stale A5 expectation is consistent with the missing reset
    branch in the supplied model excerpt. Higher P-B congestion accompanies
    worse WNS, but seed and density changed and causation is unresolved.
    
    Proposed ownership: RTL owner reviews reset sensitivity and count width;
    DV owner confirms model update paths and observation timing; STA/PD owners
    inspect full path reports and snapshot provenance. Specification owner
    defines pending-response behavior across reset.
    
    Future checks: reset/write priority, legal occupancy transition 3 to 4,
    register-model reset replay, and same-context timing/path comparisons.
    Status: no simulation, model call, or implementation run demonstrated.
    Acceptance: attach actual check results and resolve the open contract items
    before changing these planned checks to completed.

    The handoff uses role ownership because assigning work to invented people creates false coordination. It asks for a bounded decision and names evidence the reader can request. Keep the original read-only artifacts available with revision identifiers; copying only the assistant's summary loses context. If an artifact contains a command-looking line, quote it as evidence when necessary rather than presenting it as an instruction for the recipient or an agent to execute.

    Common pitfall

    False AI claim, synthetic: 'All issues are fixed and verified; congestion was the root cause, and the design is ready for signoff.' W-A contains proposals, partial excerpts, and hand calculations. None support completed fixes, causal proof, or signoff.

    Exercise and expected checks

    Rewrite this status sentence: 'AI found the bug, fixed timing, and verified reset.' Use only W-A, preserve the numerical timing evidence, and add one explicit next action. Then audit your draft for verbs such as passed, proved, fixed, sent, and approved. Each such verb needs an artifact showing that the corresponding event actually happened.

    Check yourself What is an acceptable replacement?

    Inspection identified reset-sensitivity and occupancy-width conflicts in the supplied RTL; correction proposals await review. T-B still reports -0.12 ns setup slack from required 1.85 ns and arrival 1.97 ns. Next, the RTL and DV owners should review the proposals and obtain reset and occupancy check results. This wording claims neither completed changes nor executed verification.

    Limits and reference

    A drafted note is not a sent message, an assigned ticket, or engineering approval. Attach actual results later and state which revision they cover; do not silently promote expected answers to evidence. NASA Systems Engineering Handbook, Revision 2 (2016), Appendix D verification matrix guidance provides a reference for linking requirements and verification records. The handoff format here is a synthetic engineering exercise and does not impose a particular team's approval process.

    Related topics

    • Up: Engineering Walkthroughs

    • Next: Scripts, APIs and working copies

    Previous

    71 / 173

    Next: Practical Foundations for AI Workflows

    Up next

    Up nextPractical Foundations for AI Workflows

    Sign in to save progress

    Practice this lessonReport an error

    See also

    • Specification to Requirements
    • RTL Review
    • Verification Plan and Assertions