Yurii Filipchuk 🇺🇦 -- e/acc

Founder @ Stealth AI |TechStars AI Startup Weekend SF

I build Principle, a simulation engine that stress-tests business decisions before you commit capital.

I also run partnerships and sponsorships at Techstars Startup Weekend SF and Catalyst Bay. We're a community of 10,000 founders doing IRL events in San Francisco.

Chaotic Good is my operating system. Move fast, challenge conventional thinking, put powerful tools in more hands. I believe that's how you actually make the world better. Big believer in pay-it-forward culture. If you have an idea on how we can help you or you want to help us, reach out.

The LLMs you choose say something about you. I'm team Claude for work. NotebookLM and Gemini for research. ChatGPT and Groq for personal life questions.

Previously CEO of Party.Space. Spotify was my first client, then Google, Zapier, Grammarly, UC Berkeley. Raised $2M from top European VCs.

I advise early-stage founders on fundraising and GTM. Been on both sides of the table. Invested through CYFRD: 15 deals, 2 exits.

San Francisco. Born in Kyiv.

Community Lead: Vibes & Sponsorships | Pitching Mentor

At Techstars Startup Weekend SF AI, we're hosting a fast-paced, 3-day event where tech enthusiasts come together to bring AI startup ideas to life. My role is to create an energetic and welcoming atmosphere, build connections with important sponsors, and guide participants in making their pitches clear and impactful.

Anthropic bought a Super Bowl spot to say they won't run ads.

Read that again.

30 seconds. Biggest stage on the planet. One message -- Claude stays ad-free.

No sponsored links. No paid nudges in your conversations.

This landed the same week OpenAI started testing ads in ChatGPT's free tier. "Labeled and non-sensitive topics only" -- sure. But it's still inside the interface where people share unfiltered thinking.

The business model split is now visible. Subscriptions and enterprise deals vs attention monetization. Two different incentive structures. Two different products long-term.

For anyone running strategy or workflows on AI this matters more than model benchmarks. Ad-supported AI optimizes for session length. Subscription AI optimizes for answer quality. Pick which one you want advising your board.

The privacy angle is obvious in regulated industries -- finance and healthcare won't touch an ad-funded reasoning layer. But the deeper point is trust architecture. When your AI works for an advertiser even 2% of the time you can't treat its output as decision-grade.

Anthropic made a positioning bet on the world's biggest stage.

Will it hold as they scale? That's the real question.

Founder @ Principle | Strategic Foresight Engine | Chaotic Good

This week I caught myself wondering -- do I spend more time talking to Claude or actual humans?

Then I had a call with Arvin Reddy that reminded me why real conversations still matter.

Arvin spent 11 years as Chief Technologist at Hitachi Vantara. Led AI infrastructure GTM across enterprise and government. What he said made me pull up Gartner.

The chart shows GenAI entering the Trough of Disillusionment. Dated June 2025. Seven months ago. In AI time -- ages.

Here's what hit different. Arvin's insights vs what the reports say below:

On the hype cycle

Gartner says GenAI is in the Trough of Disillusionment.

Arvin says: "It depends on the maturity of the industry itself, your vertical. Manufacturing has been doing AI for decades. Healthcare is still figuring out where to start."

On why projects fail:

McKinsey says 78% of companies use GenAI but 80% report zero material contribution to earnings.

Arvin says: "Data is basically the goldmine. The theory and discipline you apply will vary from industry to industry."

Deloitte says 93% of investment goes to technology, only 7% to people.

Arvin says: "A lot of times customers didn't know what they wanted to do. It was an education phase."

On the fix:

Gartner predicts 30% of enterprise models will be domain-specific by 2028.

Arvin was telling Hitachi this two years ago: "What if I can give you pre-trained models that get you 70-80% there? You've already short-circuited the guesswork."

On timing:

His MIT colleague told him 25 years ago: "The industry is not ready to see what this thing can do."

Reports say we're entering the "Year of Truth" where hype meets reality.

I trust the executive who's been in the trenches over the consulting deck.

The hype cycle isn't one curve. It's a thousand curves -- one per industry.

Stop asking "what ARR do I need to raise at."

It's the wrong question. Half of the Linkedin repeats it like scripture. The fact that tells you everything about how most founders think about fundraising.

SVB released their H1 2026 data. Revenue at raise for US tech companies, 2024-2025.

Median Series A: $2.5M ARR. 75th percentile: $6.5M. 25th percentile: $2.1M.
That's a 2.6× spread and nobody talks about it. Because "aim for $2-3M ARR" fits in a carousel slide. The actual distribution does not.

Seed is worse.

Median: $220K ARR. 90th percentile: $3.6M ARR.

16× spread. Sixteen. Between the middle of the pack and the top decile. But sure -- keep optimizing for some magic number you read in a thread.

Companies at the 10th percentile raise with zero revenue. Zero. Meanwhile founders with $400K ARR are in Slack groups asking "am I ready" like there's a permission slip somewhere.

There isn't.

VCs don't fund a number on a dashboard. They fund velocity. Team. Timing. Defensibility. Four things that don't fit in a benchmark table -- which is why nobody posts about them.

One more thing the data shows. The multipliers compress as you go up. Seed to Series A is an 11.3× jump at median. Series A to B is 2.4×. 

Early stage rewards outliers. Growth stage rewards consistency.

So no -- there's no correct ARR to raise at. There's only whether what you're building compounds. And if you need a LinkedIn post to tell you that, you're already behind.

The state of GenAI video no longer surprises me🤖

So we started making our own. Episode 1 of "LockedIn" is out. Who else gets frustrated receiving AWS/GCP bills?

Where should we channel the creative energy next -- something more satirical or slightly educational?

P.S. Those worms are adorable and they deserve their own show. Someone talk us out of it.

literally me when I missed a day of OpenClaw madness

Dropping a new episode of LockedIn. Inspired by a real story this time!

Hamudi Naanaa and Volodymyr Panchenko hit Anthropic's Tier 4 token limit mid-hypergrowth. 4 users to 1,400 in 3 days. 2b tokens a day! 

Their words: "A two-day response cycle doesn't match the speed we're operating at -- for us, it's 1–2 hours."

They switched to Gemini overnight. A constraint forced it.

What's the ceiling your team hasn't hit yet -- but will?

taplio

Yurii Filipchuk -- e/acc's Linkedin Analytics