GP at @MightyCapital | founder at @ProductsCount | fmr exec at @Facebook, board member at @Opera Software & lecturer at @Stanford b-school
San Francisco
Joined March 2009
- Tweets6.6K
- Following3.2K
- Followers7K
- Likes4.8K
Product discovery is shifting from persuasion to mechanics.
Buyers used to meet you through marketing pages (SEO).
Now they meet you through docs, APIs, and guides (AEO).
SEO taught persuasion.
AEO rewards clarity.
If “why it matters” isn’t embedded in the docs, you’ll be evaluated as a pile of features.
Onboarding for humans = teach.
Onboarding for agents = give access, boundaries, and a receipt.
If your product can’t say “do this / don’t do that / here’s what changed,” autonomy is a fantasy.
Most "product debates" are fights over a fake scarcity: one interface, one workflow, one answer.
That scarcity is gone 👇
Governance is the price of divergence.
Write down the “envelope” that never changes (permissions, safety, promises), then let the lane vary everything else.
Two hiring shifts I'm seeing:
Resumes ↓
Reputation ↑ (proof in a community: what you ship + how you think)
And as AI makes competence easier to fake, judgement becomes the scarce asset: what's true, what matters, what's missing, what's risky.
Think of your company like a machine.
Most days you're operating it: run the checklist, hit the targets, keep it consistent. That's tactical performance.
But periodically you have to pop the hood:
Is the machine still built for the terrain we're on now?
That's adaptable performance.
Execution compounds, but so do outdated assumptions.
If you don't schedule "pop the hood" time, it never happens.
And in a world moving this fast, yesterday's best practice becomes today's constraint.
What do you do to create space for adaptability without blowing up execution?
The most expensive failure mode in product right now isn’t “not enough data.”
It’s letting data become a substitute for judgment.
AI is making dashboards smarter and cheaper. Which means the default will be: “What does the data say?” -> ship.
That’s backwards. Data is an input. Judgment is the decision.
When data leads, a few things happen fast:
>you optimize what’s measurable, not what matters
>you stop owning the tradeoff (“the dashboard said…”)
>you lose the intuition muscle that makes great founders and product minds dangerous
2026 feels like a return to basics: AI surfaces patterns. Humans choose the bet.
The era of the "Generalist Coordinator" PM is fading.
AI feels like a return to the early mobile days where the best builders understood the silicon chips, memory, and battery. Today, the "hardware" is the model, the data, and the infrastructure.
Decisions at one layer immediately break the others, so if you don't understand the stack, you can't optimize the outcome.
The future of product management is moving from abstraction to intimacy with the stack.
After closing Fund III and our 6th IPO @MightyCapital, I’ve been spending more time talking with prospective LPs and I'm seeing a trend: LP Activism. Investors in illiquid funds are aggressively pushing for returns.
If you’re an entrepreneur backed by a firm facing this, it can have ripple effects for you: early secondary sales, unexpected liquidity conversations, or increased pressure to exit faster than you’d planned. These dynamics don’t always come from your board, they come from the board behind your board.
SC Moatti retweeted
When should companies hire their first PM? @Lovable is just hiring its first PM.
Having a PM much earlier could have helped focus the product on specific ICP and use case, which I believe is missing at Lovable.
What’s everyone’s take?
We're hiring our first product managers at Lovable.
It's taken us a while to figure this role out:
1. Bias to ship. You should love an environment with urgency. Remove blockers, scope down ruthlessly, and ship.
2. User-obsessed. Talk to users and bring findings with conviction and rigor. Know what's working, what needs iteration, what should be killed.
3. Technically deep. Engineers should never feel like they're getting directions out of context. You've coded before. You build prototypes yourself.
4. Excellent product taste. You question things deeply and have strong opinions about where software is going.
See the role here: lovable.dev/careers?ashby_ji…
90% of roadmap planning is just organizing noise.
AI is finally killing the "backlog tetris" that distracts founders from actual company building.
We are seeing the best teams go back to basics. Ignore the Cash Cows (let the AI sustain them) and obsess over the Rising Stars.
Less grooming. More betting.