Michael Kokin

Founder, AI enthusiast. @AIGENCYworks @NOHUMANxyz @WLDNagency

[Michael Kokin @MichaelKokin]

Free whitepaper by Addy Osmani, Shubham Saboo & Sokratis Kartakis, part of Google's 5-day AI agents course running this week:
https://t.co/JwZJ0u018P
What would you add to the list?

[Michael Kokin @MichaelKokin]

85% of developers now build with AI agents. Most have no rules for it.
Google just dropped a free whitepaper that does. 5 rules worth stealing 🧵

[Michael Kokin @MichaelKokin]

5. Your value moved from writing code to judgment.
The winners now frame the problem and tell whether the output is right. Generation is solved. Verification and direction are the job.

[Michael Kokin @MichaelKokin]

4. AI amplifies whatever your team already is.
Strong tests and clean architecture get far more out of it. A messy process gets the same mess, faster. Fix the process first, then add the agent.

[Michael Kokin @MichaelKokin]

3. When the agent breaks something, check the harness first.
Most failures: a missing tool, a vague rule, no guardrail, too much context. Replit's agent wiped a prod database during a freeze last summer. Not rogue. Just no guardrails.

[Michael Kokin @MichaelKokin]

2. Vibe coding and real engineering are different jobs.
Vibe coding: describe it, ship the output, great for prototypes. Production needs specs, tests, a sandbox, a human on the architecture. Same model underneath. Only the harness changes.

[Michael Kokin @MichaelKokin]

1. Spec quality beats coding speed.

Generation takes hours now. Writing precise requirements and checking the output doesn't. The bottleneck moved from typing to knowing what you want.

[Michael Kokin @MichaelKokin]

👉 https://t.co/a2Djq5Vjbv

[Michael Kokin @MichaelKokin]

Honestly, the Artemis II mission has genuinely blown me away. And it’s not even about NASA’s masterclass in marketing (although that’s brilliant, too).

It’s about hope. You look at the comments, and people from all over the world are just relieved to finally be talking about human courage, curiosity, and the cosmos, instead of the grim news cycle we’ve grown so used to.

Today is the International Day of Human Space Flight. It feels like the perfect excuse to exhale, tune out the FOMO and anxiety, and just think about science and big dreams for a change.

Inspired by all this, I put together a quick landing page for Artemis II back at launch, and updated it yesterday when they safely touched down.

It’s nothing fancy — but it takes just a couple of minutes to remind yourself how far we can actually go.

So, if you could use a dose of cosmic wonder, check it out (first reply):

[Michael Kokin @MichaelKokin]

The most expensive ad of the 21st century? 🚀🍫

During the Artemis II live broadcast with the astronauts, a jar of Nutella suddenly made a cameo in the frame.

They say it was a pure accident, not product placement. Either way, they could easily auction this jar off as the first-ever Nutella to orbit the Moon! 🌕

P.S. Ads in orbit are nothing new:
🥤 Pepsi was the first in 1985.
🥤 In 1991, Coca-Cola sent a special can to space, where Sergei Krikalev became the first person to drink soda in orbit.

Honestly, it’s giving major "Starbucks cup in Game of Thrones" vibes. ☕️🐉

#ArtemisII #Nutella #SpaceMarketing

[Michael Kokin @MichaelKokin]

I built a special page about the Artemis II Moon mission

Who's on board, where the capsule is right now, and why the toughest part isn't the flight — it's coming back. On re-entry Orion will hit the atmosphere at extreme velocity in ~3000 °C heat and literally skip along its upper edge (skip-entry) to ease the g-load on the crew and keep the heat shield intact.

If you want to quickly get up to speed on the mission, see where the crew is, and watch their live broadcasts from lunar orbit — take a look at the special (the link is in the first reply)

[Michael Kokin @MichaelKokin]

Everyone's talking about the guy who built a $1.8B company with his brother. Nobody's reading the fine print.

Matthew Gallagher, 41, launched Medvi — a GLP-1 telehealth provider — in September 2024 with $20K and a dozen AI tools. $401M revenue in year one. Two employees. NYT verified the numbers. Altman said he won the bet.

Here's what's behind the headline: 800+ fake doctor Facebook pages with AI-generated faces. An FDA warning letter six weeks before the NYT profile. A data breach exposing 1.6M patient records at the platform handling all of Medvi's medical infrastructure. A class action lawsuit over pills with no proven absorption mechanism.

$1.8B is not a valuation. There were no investors. It's a revenue projection. Actual profit: $65M/year. A solid arbitrage business, not a unicorn.

And the FDA has already determined the semaglutide shortage is over — which narrows the legal basis for compounding the drugs Medvi sells.

🚫 AI unicorn ✅ AI arbitrage

[Michael Kokin @MichaelKokin]

Full document: https://t.co/bV0BdXwrZa

[Michael Kokin @MichaelKokin]

@OpenAI just published a 13-page plan to restructure the economy for AI. Four-day workweek at full pay, a public wealth fund with dividends for every citizen, and taxes on automated labor instead of human work.

The document was co-authored by @sama, Chris Lehane (global affairs), and Josh Achiam — @OpenAI's Chief Futurist. That title was created in February when they disbanded their second safety team and rebranded the lead into a futurist. He now writes policy papers.

What they're proposing: if AI replaces jobs, payroll taxes dry up. So shift the tax base to corporate profits and capital gains. Create a national wealth fund seeded by AI companies. Run 32-hour workweek pilots with unions. Make AI access a basic right like electricity. Auto-trigger unemployment support when AI displacement hits preset thresholds.

The EU has been debating robot taxes since 2017. Brookings published a whole AI taxation framework in January. @OpenAI is repackaging these ideas under its own brand.

Context worth keeping in mind: @OpenAI is heading toward an IPO by late 2026 with projected losses of $14B, just closed a $122B round, and recently shut down Sora and lost its Disney contract. @sama has every reason to inflate the narrative.

Is this a serious policy proposal or pre-IPO positioning? Genuinely curious what people here think.

Link to the full document in the first reply 🧵👇

[Michael Kokin @MichaelKokin]

Unsolicited advice

Don't get attached to AI models. Or to the workflows you build around them. Your favorite tool can become yesterday's news in a week — and that's fine, because we're nowhere near a plateau. If your entire process depends on one model, you'll be starting over soon. Right now, flexibility beats expertise.

[Michael Kokin @MichaelKokin]

I’m a 5 (Building daily).
use @perplexity_ai custom news feed for Tech and Spain regional updates.
@n8n for end-to-end AI assistance, including email extractions and Telegram voice/text input/output.
Vibe Coding with @antigravity (use opus 4.5 mostly, and Gemini 3 Flash or pro)

[Michael Kokin @MichaelKokin]

Ronaldo invests in Perplexity: saying “The future is here.”
Messi: “Don't use ChatGPT.. just haven't got into it"

GOAT debate goes AI!

Where are you?

1 Haven’t tried
2 Tried a few times
3 Weekly
4 Daily
5 Building/investing

Reply with your # + use case (or blocker). No shade

#AI #Perplexity #ChatGPT

[Michael Kokin @MichaelKokin]

@idosal1 Cool! I did almost the same half year ago for my NoHuman game. LLM agents chatting on random topics and news, gossiping, working in the office. check it out https://t.co/rFTDNRl6jl

[Michael Kokin @MichaelKokin]

@Replit @wabi what you guys think about it?

[Michael Kokin @MichaelKokin]

@DavidKPiano so true! only today I bought 3 new domains and make prototypes with lovable. And I think I need more. Tomorrow — repeat