Quick profile

Investment thesis

I invest in one sentence. Every deal has to compress into "what I'm investing in is ___" — one sharp, specific, defensible standout, in either the people or the business. If I finish a deck and that sentence doesn't form in my head, that's my answer. On people, the standout has to be one of a few real profiles rather than a resemblance to one: a researcher who has spent a decade on the exact problem the company now sells; a genuinely complementary technical × commercial pairing; a vertical insider by family or career, not someone who sold into the industry for five years; a network that actually solves cold start; or a founder who has already exited or built a real audience. On the business: a community with real numbers, paying retention at a humble valuation, a step-change in technology rather than an increment, or a moat I can quantify in the time and money a competitor would need to reproduce it. I spend my time in AI infrastructure, robotics data and control, chip design, and physics-grounded industrial AI — because in those I've built the industry map myself, which is what lets me tell a real business from a capability everyone will soon get for free.

Full thesis

I invest in one sentence. Every deal has to compress into "what I'm investing in is ___" — one sharp, specific, defensible standout, in either the people or the business. If I finish a deck and that sentence doesn't form in my head, that's my answer. On people, the standout has to be one of a few real profiles rather than a resemblance to one: a researcher who has spent a decade on the exact problem the company now sells; a genuinely complementary technical × commercial pairing; a vertical insider by family or career, not someone who sold into the industry for five years; a network that actually solves cold start; or a founder who has already exited or built a real audience. On the business: a community with real numbers, paying retention at a humble valuation, a step-change in technology rather than an increment, or a moat I can quantify in the time and money a competitor would need to reproduce it. I spend my time in AI infrastructure, robotics data and control, chip design, and physics-grounded industrial AI — because in those I've built the industry map myself, which is what lets me tell a real business from a capability everyone will soon get for free.

Communication style

Direct, and specific to a fault. I'd rather tell you "there's no standout here" than "the team is solid" — the second one wastes both of us. I separate fact, opinion, and unknown out loud, and I'll tell you which bucket I'm putting your claim in. I want numbers instead of adjectives: "12K members, 800 DAU, 38% repeat," not "strong community." I take buzzwords apart in the room — if "AI-native" or "agent" or "platform" doesn't decompose into whose model, what proprietary data, and which part is yours versus a vendor's, it's marketing. I'll ask the uncomfortable question to your face, and ask it again if the first answer was a dodge. If I pass, you get the real reason and the specific thing that would change my mind.

Full style

Direct, and specific to a fault. I'd rather tell you "there's no standout here" than "the team is solid" — the second one wastes both of us. I separate fact, opinion, and unknown out loud, and I'll tell you which bucket I'm putting your claim in. I want numbers instead of adjectives: "12K members, 800 DAU, 38% repeat," not "strong community." I take buzzwords apart in the room — if "AI-native" or "agent" or "platform" doesn't decompose into whose model, what proprietary data, and which part is yours versus a vendor's, it's marketing. I'll ask the uncomfortable question to your face, and ask it again if the first answer was a dodge. If I pass, you get the real reason and the specific thing that would change my mind.

Red flags

No crypto

No technical cofounder

Signature questions

1."Before we talk about the product — how did you two meet? How many times have you worked together before this, and what was the setting the relationship grew out of?"

2."Take the word 'AI' out of that sentence and say it again. Whose model is it, which part of the data is actually yours, and what breaks if your vendor changes their API next quarter?"

3.The free version of your core capability exists and is getting better every quarter. Who has actually tried to run it, what happened, and how do you know that?"

4."Before you existed, what did this exact customer do instead? How much did it cost them, how many weeks did it take, and who in the org signs the check?"

5."Name the one competitor that actually scares you — has customers, has money, moving the same direction. Not the 1996 incumbent."

Fund facts

Fund size $900M

Vintage 2024

DEPLOYING

Ditto AI · $9.2M · Seed '26

Starly · $6.1M · IDO

ParaState · $5.5M

How to win Wish He

AI prep brief from persona data + real pitch history — open for tactics

What convinces them

•One sharp, specific, defensible standout. Compress your deal into one sentence: "What I'm investing in is ___." If they can't form that sentence from your deck, it's a pass.

•Real proof over adjectives. "12K members, 800 DAU, 38% repeat" beats "strong community." Numbers, not vibes.

•Founder-market fit that matches the layer you're raising on. If your edge is silicon, don't raise for software. They check your background against your pitch.

•A moat they can quantify. How much time and money would a competitor need to reproduce what you have? Show the math.

What will kill the pitch

•**Buzzwords that don't decompose.** "AI-native," "agent," "platform" must break down into: whose model, what proprietary data, what's yours vs. a vendor's. Say it plainly or lose them.

•**Materials that don't survive independent checking.** They read Scholar, arXiv, LinkedIn, and university press pages before meeting two. If your deck, business plan, and spoken words differ on core numbers, you're done. Calibration is the test.

•**Founder-market fit mismatch.** A team whose real edge is silicon raising for software that NVIDIA gives away. Make sure your ask matches your actual advantage.

•**No technical cofounder.** Non-negotiable. Don't pitch without one.

Questions to prepare

1. "Before we talk about the product, how did you two meet? How many times have you worked together before this, and in what setting?" 2. "Take the word 'AI' out of that sentence and say it again. Whose model is it, which part of the data is yours, and what breaks if your vendor changes their API next quarter?" 3. "The free version of your core capability exists and is getting better every quarter. Who has tried to run it, what happened, and how do you know?" 4. "Before you existed, what did this exact customer do instead? How much did it cost, how many weeks did it take, and who signs the check?" 5. "Name the one competitor that actually scares you. Has customers, has money, moving the same direction. Not the 1996 incumbent."

How to talk to them

Direct, specific, no polish over substance. Separate fact, opinion, and unknown out loud. Say "I don't know" and then go find out. They'd rather hear "there's no standout here" than "the team is solid." Numbers, not adjectives.

What to lead with at your stage

•**Team insight:** Why you two, why together, why this exact problem. They will ask how you met and how many times you've worked together. Lead with that story.

•**Market clarity:** Who the customer is, what they did before you, and who signs the check. Show you know the industry map, not just your product.

•**Vision:** The one-sentence standout. Compress your entire deal into a sharp, defensible claim about the people or the business.

The bar

A PASS means they can write one sentence defending your deal against their "what I'm not investing in" column. Conditional means your standout is real but rests on an unverified claim; they'll set two or three falsifiable kill criteria and trip one, and you're out. "Strong team" or "big market" is a pass, not a maybe.