Ilya Romanov

Marketing

[Ilya Romanov @ilyshka]

@maubaron @maubaron how much max would you recommend spending on testing each individual video with Spark Ads before deciding to kill?

[Ilya Romanov @ilyshka]

Gamma hit $100M ARR without turning into an ad machine.

Most startups would’ve done the opposite.

That’s why their marketing feels like a slap in the face to the usual “let’s pump traffic and figure it out later.” 👇

1️⃣ They didn’t confuse hype with PMF.

Gamma won Product of the Day/Week/Month on Product Hunt—and still admitted the product wasn’t good enough.

Instead of pouring money into ads, they spent 3–4 months of the whole team on the first 30 seconds of onboarding.

2️⃣ Their “secret channel” isn’t Facebook Ads.

It’s word of mouth + branded search, which still drive more than half of new signups.

Paid is an amplifier, not a crutch. If paid >50%, for them it’s a red flag—not “normal, that’s how everyone does it.”

3️⃣ Influencers: no “let’s hire three celebrities and win.”

Gamma bets on thousands of micro‑influencers: teachers, consultants, educators, creators.

Grant Lee personally onboards them on calls. Yes, the founder actually jumps on Zoom with creators and explains the product.

4️⃣ They opened their brand to the world.

Instead of guarding the brand like a sacred asset, they built an open brand hub so any creator can easily produce on‑brand content.
The brand isn’t a brake. It’s a growth accelerator.

5️⃣ They don’t worship a “single best channel.”

Word of mouth, micro‑influencers, LinkedIn, performance, onboarding experiments, AI models, pricing—everything orbits one question:

“How do we get one more user to tell one more person about us?”

In a world where most startups start with “what’s our CTR?”, Gamma started with “are the first 30 seconds of our product magical enough to deserve traffic at all?”

This post is based on insights from Lenny’s Podcast conversation with Grant Lee.

If you’re interested, I can share the full video link.

[Ilya Romanov @ilyshka]

Most teams still sell AI as “automation”.
The data says users want something else

A 1.5M‑conversation study on ChatGPT shows:

- 49% of messages are Asking (guidance, advice, explanations)
- 40% are Doing (tasks, content, code)

“Asking” is growing faster and gets higher satisfaction scores

People don’t just want AI to do work for them.
They want AI to help them decide what’s worth doing at all

If your AI product is positioned only as “automation”,
you’re probably optimising for the shrinking half of usage—
and ignoring where users feel the most value: decision support in knowledge work

[Ilya Romanov @ilyshka]

Sora is the queen of demo reels.

Kling is the workhorse of real production.

From talking to 15+ AI creators, here’s what actually happens in their workflows:

No one is rendering recurring client videos with Sora.

Everyone is running a conveyor belt: GPT → MidJourney → Nanobanana Pro → Kling.
Why?

Sora’s access and limits = pain for production teams.

Kling is cheaper, more predictable, and much easier to plug into an existing pipeline.

The market has quietly voted already:

“Let Sora shine in launch demos. We need tools that survive 100+ videos a month without breaking.”

Want the real insider take?

Right now the money isn’t going to the people with the prettiest model demo.
It’s going to the people whose stack doesn’t fall apart under load.

[Ilya Romanov @ilyshka]

https://t.co/JUhG11lfqB

[Ilya Romanov @ilyshka]

“95% of AI pilots fail”

Sounds bad, until you look closer

In my new article at https://t.co/KmycixI9if I explore why those numbers might actually be good news for the future of AI deployment

👇👇👇

[Ilya Romanov @ilyshka]

Back here after such a long break it probably looks like I’ve never posted at all😅

Tomorrow I’m sharing my new piece on MIT’s NANDA report (July 2025), why a 95% AI pilot failure rate isn’t the end, its the path to impact

Like this so it actually shows up in your feed tomorrow

[Ilya Romanov @ilyshka]

@Blastup_io @Blast_L2 to the moon!

[Ilya Romanov @ilyshka]

@withhimu tribe

[Ilya Romanov @ilyshka]

10/10 If Paige's story struck a chord or taught you something, please like and retweet the first tweet. Let's share these lessons to help others in the tech community! 🚀 #TechCommunity #sharingiscaring

https://t.co/bQhT2F3FZ5

[Ilya Romanov @ilyshka]

1/10 Ever thought about starting your own AI startup with all the AI hype around? It's not always a smooth ride. Let's dive into Paige's story, an AI website builder, and learn from its challenges. 🧵 #AIStartup #RealTalk

[Ilya Romanov @ilyshka]

1/10 Ever thought about starting your own AI startup with all the AI hype around? It's not always a smooth ride. Let's dive into Paige's story, an AI website builder, and learn from its challenges. 🧵 #AIStartup #RealTalk

[Ilya Romanov @ilyshka]

9/10 Nick Naumov's story with Paige is a real-deal look at the ups and downs of starting a tech business. It's full of lessons for anyone thinking about jumping into the startup world. #StartupJourney

[Ilya Romanov @ilyshka]

@sealaunch_ Got you, thanks!

Still would be interesting to know if the btc transactions on Okx should be on dApp Radar stats

[Ilya Romanov @ilyshka]

@DappRadar Hey team, quick check – are we sure about OKX NFT Marketplace being in the top 2 by volumes for the past 30 days? Seems a bit off, especially with their significant NFT volume coming from BTC transactions. Can you guys look into this? Thanks!