QWalk me through how you started your last project in Claude Code.
AI had Codex map the repo and draft a plan, then rewrote the auth parts myself.
VP Product · Meridian Labs
QWhat would make our new diff-review feature something you'd actually use?
AGroup changes by intent, not by file — I lose trust when one change is scattered everywhere.
Staff Engineer · Platform
QWhen did an agent last fail you, and what did you do?
AIt looped on a failing test, so I re-ran the prompt twice and finally fixed it by hand.
AI Product Engineer
QWould you let an agent merge to main on its own?
ANever blind — but a staged preview I can approve? I'd adopt that tomorrow.
Technical Founder
Install the AskReal skill in Claude Code or Codex. It records how you actually work — prompts, tools, and decisions.
$ askreal capture
✓ Scanning Claude Code sessions…
✓ 127 sessions · 84 Codex sessions
✓ Tools: Claude Code, Codex, Cursor
✓ Workflows mapped: 19
→ Ready to upload
Push only the sessions you choose to share. Keys, client code, and anything private are stripped first.
A short voice interview fills the gaps your sessions can’t show — the tools that frustrate you, the workarounds you’ve settled on, and what you wish worked better.
Your approved workflows and interview answers become an AI avatar that can explain how you use AI — grounded in real data.
VP Product · Meridian Labs
San Francisco · 11 yrs · B2B SaaSOwns the roadmap for a ~180-person B2B SaaS company, balancing enterprise requests against longer-term platform bets.
Has shipped IDE plugins, CLIs, and API platforms, so she judges AI tools through a builder's lens rather than a buyer's.
Sets the team's AI-tooling standards and reviews how each PM folds agents into discovery and delivery.
Ran the rollout that moved four engineering squads from ad-hoc Copilot use to structured Claude Code workflows.
Reaches for Claude Code on anything she'll ship, and Codex when she just needs to understand a repo fast.
Available for live interviews during PT business hours, and async for everything outside them.
Saw a dev-tools product through acquisition, so she has firm views on what makes tooling sticky versus disposable.
Still reads and reviews diffs herself, which shapes how critically she evaluates agent output.
Hosts a biweekly forum where engineers trade prompts, failures, and workflow wins.
Fluent enough in three stacks to tell whether an agent's changes are actually idiomatic.
Loved the chat with the avatar? Ask for a meeting with the real person. When you want to go deeper, request time with the practitioner behind it — pick a slot and book a live session or product test.
REQUEST A MEETINGFOR CLAUDE CODE AND CODEX USERS
Import the workflow data you choose to share. Approve the AI twin it builds, earn when companies interview it, and accept live conversations only when the opportunity is worth your time.
CONTRIBUTE YOUR WORKFLOW