Hossein

I help overworked professionals turn their job experience into a startup without quitting their job or learning to code | Now building Humanic | Ex. Amazon

[Hossein @tootoonchy]

jev is now publicly available. Can’t wait to build some cool features with it. What are you going to use it for ?

[TypeSafe AI @typesafeai]

Jev is now available to everyone. No waitlist.
Start using it here: https://t.co/5MfN4rIif5

[Hossein @tootoonchy]

The message sitting in your drafts folder is not a strategy.

Week 5 of my program for employed founders is one assignment: write one warm-network message tonight. Not a deck. Not a funnel. One real message to one real person.

Here's what makes it work:

Name their context. "I saw you moved to VP at [Company] last month." Ten seconds of homework. That one line shows you're not blasting a list.

State one problem. Not your whole offer. Not your three packages. The one specific problem you know how to fix.

Make the ask small. "Would a 20-minute call be useful?" is not a commitment. It's a question.

No attachments. No links. No pricing. A first message loaded with files reads like a sales packet. It gets archived.

Send it before you perfect it. The message you edit for two weeks converts no one.

The warm-network message that gets a reply is specific, short, and actually sent.

Your first client is not at the end of a funnel. They're already in your phone.

If you're building a startup while employed and want the full Week 5 messaging framework, the community is free: https://t.co/Um9gnmFm79

Program details at https://t.co/zrhDZ5ifJD

Who in your network should get that message tonight?

[Hossein @tootoonchy]

@kybr_dev using podcast instead of writing blogs. Podcast is the only format that checks all Google guidelines as non-commodity. So does video but that is too much work.

[Hossein @tootoonchy]

How to get your business cited by AI search engines https://t.co/OCDR22nYXP

[Hossein @tootoonchy]

Humanic : https://t.co/OyKkG7Bu7C

[Hossein @tootoonchy]

Google Guidelines : https://t.co/3cYXi1eVlU

[Hossein @tootoonchy]

The price objection is never really about price.

Most founders hear "too expensive" and go straight for the wrong fight:

Explaining ROI

Benchmarking against competitors

Offering a discount to close

None of that works when the buyer has never felt the problem.

Dr. Sagar Bansal knows this better than most. When he was building his $100 smart blind cane, investors kept pointing at $3 aluminum pipes and asking why anyone would pay more.

His answer stopped the room.

"This is not a demand problem. It is a supply problem."

People who have never lived with daily fear will always under-price the thing that removes it. They are not comparing products. They are comparing two objects they have never needed to trust with their safety.

Sagar did not argue price. He handed a VC a blindfold and asked them to walk across a plain surface with a single step in the path.

Then he asked them to imagine that every single day.

The price conversation changed immediately.

That is lived experience transfer. It is the only thing that actually works before a price debate.

Before you try to defend your number, ask yourself: has this person ever felt the problem your product solves? If not, no spreadsheet will get them there. Put them in the scenario. Run the demo that makes them feel the gap. Let the constraint do the convincing.

Once they have lived it, even briefly, they stop comparing your product to the cheap substitute that only looks the same.

I spoke with Sagar on a recent episode of Humanic about how he navigates investors, boardrooms, and high-stakes decisions without sight. The full conversation is worth your time.

https://t.co/5wUKlB6GUJ

https://t.co/5uU4xe81PK

[Hossein @tootoonchy]

Most teams are interrogating AI wrong.

When a predictive model makes a decision, the instinct is to paste it into ChatGPT and type: “Why did this happen?”

That question, asked that way, is a trap.

Afrooz Ansaripour runs data science and AI at Walmart. She told me:

Don’t ask a blank LLM why. Feed it a structured fact packet first.

What features contributed to the decision? What was the confidence score? What business rules were in play? What’s the relevant context?

Then ask GenAI to translate those facts into plain language.

That’s a completely different job. Translation, not reasoning. The LLM becomes a communication layer, not a new logic layer.

When you skip the fact packet and let the model reason on its own, it invents an explanation. A smooth-sounding one. One that has nothing to do with what actually happened inside your system.

The explanation sounds right. It just isn’t.

Keep explanation separate from decision-making. Your predictive models make the call. GenAI only explains it after the fact.

Two separate jobs. Two separate systems. One clean line between them.

We unpacked this and a lot more on the latest Humanic episode. If you’re building AI systems that real humans have to trust, it’s worth your time.

https://t.co/5uU4xe81PK

[Hossein @tootoonchy]

Filing a patent is not permission to ship.

A patent lets you exclude others. It does not clear you to make, use, or sell.

Tom Joseph’s skin-lotion example: two valid patents, one inventor still blocked.

Before you treat “we filed” as a green light, ask what still sits between the claim and a lawful product.

https://t.co/cMXZAqE1vs

https://t.co/5uU4xe81PK

[Hossein @tootoonchy]

Founders who marry their first idea divorce it expensively.

[Hossein @tootoonchy]

Federated learning just killed the "you have to share your data" argument in pharma.

It reframed it entirely.

The standard pitch for cross-organizational AI goes like this: pool your datasets into one place, train a shared model, and everybody benefits. The problem is that pharma companies treat proprietary molecular data like nuclear codes. That pitch doesn't work. It never did.

Arash Atashnama, co-founder of https://t.co/0oA4ObuznY, takes a different starting position: ask whether the learning needs to be centralized at all.

With federated learning, it usually doesn't. Models train across distributed environments. Sensitive data stays closer to where it lives. No central warehouse, no single point of exposure.

Arash is careful about where the real work begins. "Trust is the key," he said. "Companies need governance, auditability, and confidence that their IP is protected. So the technology matters, but the trust layer matters just as much."

This is the part that separates a working system from a compelling slide deck.

If you bring an algorithm to a pharma partner without a governance story – without auditability baked in, without a clear answer to "what happens to our IP" – you do not have a shareable system yet. You have a technical proof of concept that needs further development to become a solution.

Consider this question: in your own organization, when you ask teams to collaborate on sensitive data, are you leading with the technology or the trust structure?

Usually it's the technology. Usually that's why it stalls.

Full conversation with Arash Atashnama – Silos, synthetic data, and trust: How federated learning unlocks pharma's 80% dark data.

https://t.co/Pgt48uNpC9

https://t.co/5uU4xe81PK

[Hossein @tootoonchy]

An executive's job is different. Your job is to hold the original goal in view and notice when reality has drifted from it.

You can't do that from inside the problem.

When did you last step far enough back to see the whole phone?

https://t.co/o6Qun06hzw

[Hossein @tootoonchy]

Most founders I've met don't lose because they made the wrong decision.

They lose because they were too close to see they were drifting.

Dr. Sagar Bansal shared something on Humanic that I keep thinking about.

He has a small visual frame. When his phone slips outside it, he doesn't reach for the phone. He steps back until the whole phone fits inside what he can see.

That's it. One move.

[Hossein @tootoonchy]

He applies the same move to every company decision. When something feels unclear or off, he doesn't dig in further. He steps back. He compares where the company started with where the company was supposed to go, and he corrects the deviation before it compounds.

Founders, by nature, go the other direction. They go deeper. They become the engineer. They optimize the broken thing instead of asking whether the broken thing is even the thing to build.

[Hossein @tootoonchy]

Pasting your invention into ChatGPT to draft a patent may have already killed it.

You disclosed it publicly.

Pasting invention details into a public generative AI tool can count as a public disclosure. In some countries, that alone is enough to invalidate your patent application before you've filed it.

The US gives you a one-year grace period after first public disclosure. That sounds like breathing room. It isn't.

It doesn't fix your foreign filings. If your invention is still being developed, you may be foreclosing future versions of it you haven't even built yet.

I talked to Tom Joseph about this recently. He's spent years watching founders assume a fast AI-generated draft is a strong patent. His point that stuck with me: the danger isn't the hallucinations. It's the disclosure risk you didn't know you'd already triggered, and the gap between what founders think they filed and what they actually protected.

Speed and strength are not the same thing. A patent that's quick to file and impossible to defend is worse than no patent – you've paid for a false sense of protection.

Before you paste anything into a public AI tool, ask yourself: would you be comfortable if your competitor read this right now?

Full conversation with Tom: https://t.co/cMXZAqE1vs

https://t.co/5uU4xe81PK

[Hossein @tootoonchy]

This is @HumanicSpace

I built this to help experts, coaches, consultants founders, and operators share their expertise with the world and allow their customers to discover them.

We are releasing a big update with exciting features to rank your business #1 on AI search engines.

If you want to be the first to try and secure your username, Please like this post and send me a DM to get started.

[Hossein @tootoonchy]

The quietest communities are sometimes the healthiest ones.

That's not obvious. It wasn't obvious to Tayla Tulloch either, until she caught herself doing something that quietly undermined everything she was trying to build.

She was adding more prompts, more challenges, more structured participation. The more she pushed, the more it felt like she was holding something together that wasn't learning to stand on its own.

Then she named it: "I was managing activity instead of nurturing culture."

There's a real difference between those two things.

When you manage activity, you become the center of every interaction. Members show up for you, not for each other. The moment you stop prompting, the momentum stops with you, because the community never learned to move without you at the center.

When you nurture culture, something different happens. Members start finding each other. Connections form that have nothing to do with you.

Tayla put it plainly: "Silence doesn't always mean disengagement, and constant noise doesn't always mean connection."

A quiet week might mean members are building trust in the background, not drifting away. A busy week might just mean people are responding to prompts without ever actually connecting to each other.

The healthiest communities Tayla has seen aren't built around one person carrying the culture. Members create connection and momentum with each other. The manager's job is to build conditions where that can happen, then trust the community to move.

If you're over-prompting and over-checking, you might not be growing your community. You might be preventing it from learning to grow on its own.

Consider whether you are managing activity, or nurturing culture.

Full conversation with Tayla Tulloch: https://t.co/Eh6XJwFh4I

https://t.co/5uU4xe81PK

[Hossein @tootoonchy]

Neil Hoyne has the full breakdown here, and you can try the tool directly at https://t.co/RM4Od9Qha5: https://t.co/RFmXEv0Hat

[Hossein @tootoonchy]

Google released a new tool today called Pics, and it fixes the most annoying thing about AI image generation.

[Hossein @tootoonchy]

What I find interesting is the strategy. Google didn't try to beat Midjourney on raw image quality. They watched how people actually use these tools and built around that workflow instead.