AI

AI used well is a force multiplier for DV engineers; AI used badly is a liability that ships latent bugs. This page is a map of where AI fits in the RTL and UVM workflow — prompt and context engineering, pair programming, debugging, code generation, agentic patterns, and the honest limits — with each card opening into a deep-dive post.

The foundational reference is the Practitioner Playbook below. The remaining cards drill into specific themes — how to write better prompts for RTL or DV, how to use AI as a real debugger, how multi-agent systems like HAVEN and UVM² achieve their published coverage numbers, and where the current research draws the line.

Foundations — Prompt & Context Engineering

The shift the field made in 2025: from optimizing one prompt to engineering the context, tools, and reasoning loop. These cards are where every other technique starts.

📖

AI Playbook for DV

Prompt, context, pair programming, debug, code gen, agentic patterns — the full reference

✍️

Better Prompts for RTL

Spec-as-prompt, few-shot from your codebase, CoT for timing and protocol reasoning

🧪

Better Prompts for DV

UVM component scaffolds, sequence libraries, coverage prompts, decomposition patterns

🪟

Context Engineering

Minimum viable context, tool-per-concept, scratchpads, stale-context pruning

AI as Collaborator

The strongest empirical results in AI-for-code live in debugging and scaffolding. These cards cover the day-to-day patterns.

👥

AI as Pair Programmer

UVM scaffolding, refactor proposals, commit explanations, rubber-duck debug

🔍

AI as Debugger

Hypothesis-rank, CoT on log slices, self-debug loops, invariant-violation checks

📝

Documentation & Explanation

Auto-generated component docs, legacy-code walkthroughs, RTL behavior explanations

Code & Verification

Where AI-generated code actually wins, where it loses, and how to keep it from shipping silently broken work.

⚙️

Code Generation Limits

What wins (scaffolds, repetition), what fails (novel RTL) — CVDP 34% ceiling research

💳

Tech Debt in AI Code

Model-stack workaround, model dependency, performance-optimization debt

✅

Validation Gates

Compile + lint + smoke gates for every AI-generated UVM and RTL artifact

Agentic Systems

The 2026 research frontier — agents that reason, act, self-critique, and orchestrate. With published UVM systems hitting real coverage numbers, the fusion is no longer hypothetical.

🤖

Agentic Design Patterns

ReAct, Reflexion, Plan-and-Execute, multi-agent specialization for DV

🎭

Multi-Agent UVM

HAVEN, UVM², UVMarvel — published agent systems for UVM synthesis

🔁

ReAct & Reflexion Loops

Reason-act cycles and self-critique loops applied to triage and codegen

Limits & Research

The honest version — where AI in DV still does not work, and the literature you should read to keep current.

⚠️

The Honest Limits

Multi-cycle RTL, novel protocols, long-context degradation, the 34% ceiling

📚

Research Reading List

30+ arxiv papers from 2024-2026 grouped by theme

Start Here

1. Read the AI Playbook for DV — the foundational reference covering all themes on this page.
2. Pick one debugging pattern from the "AI as Debugger" card and try it on your next failing test — hypothesis-rank is the lowest-friction entry.
3. Add a validation gate to any workflow where AI-generated code lands in your codebase — compile + lint + one smoke test minimum.

For the debugging-specific deep-dive map see the Debug page. For UVM design patterns these techniques compose with, see UVM.