CEO at Vermillion Sky | Mobile App Expert: Helping Founders Build Apps with Positive Revenue & Cash Flow | Father of 3 | Gamer

Most mobile app projects fail for boring reasons: weak product thinking, fuzzy UX, and no clear line to revenue. Not bad ideas.

I run Vermillion, a mobile product and app development studio that helps PE portfolio companies and funded startups design, build, and ship flagship apps. React Native, iOS, Android, backend – from first sketch to shipped product.

Our core work:

Product & UX – discovery, product strategy, user flows, and interface design

Engineering – React Native, native integrations, APIs, and infrastructure that doesn’t fall over at scale

Monetization baked in – subscriptions, in‑app purchases, paywalls, and billing flows designed into the product from day one

Who we’re best for:

PE Operating Partners modernizing legacy platforms into mobile-first tools their teams actually use

Founders who just closed a round and need a real product in-market, not another deck

Over the past 15+ years I’ve helped multiple founders launch, scale, and exit, including several with strong revenue and positive cash flow at time of sale. One example: PicklePlay, a club management platform we helped take from idea to acquisition.

I take on a small number of builds each quarter so I can stay close to product, not just sit in status meetings.

If you’re a PE-backed operator or funded founder and your next move is a serious mobile or web app let’s talk.

📩 ken@vermillion.agency

🔗 vermillion.agency

Ken Vermeille

Legacy code made ai replicate our worst architectural decisions

We're now at the point where every project uses AI.

Every single one.

And it's working. 30% productivity increase across the board.

But here's what I'm wrestling with right now:

Traditional estimation is dead.

Our Fibonacci sequence (1, 2, 3, 5, 8, 13) for task sizing? Out the window.

A "3" used to mean a day and a half. A "1" meant an hour.

Now? Who knows.

Some tasks AI crushes in minutes. Others still take the full time.

There's no pattern yet.

The idea of sprints is dead too.

Two-week cycles. Sprint ceremonies. All those hours of coordination.

They existed to keep everyone on the same page.

But when AI can instantly sync with your codebase, instantly understand the context?

What replaces it? I don't know yet.

Something will emerge. We're in that weird in-between phase.

Here's what I do know:

We're heading toward a world where developers become magicians.

All those internal tools you always wanted to build? You can now.

All those side projects that sat in the backlog forever? Possible.

But one thing AI will never reproduce:

Taste. Architectural decisions. Real-world consequences.

Humans live in real life. We see and feel the impact of our choices faster than any AI.

That's not going away.

So yes, we're moving fast. 30-50% faster on most projects. Sometimes 80%.

But we're also more disciplined than ever.

Because speed without standards is just technical debt at scale.

And I'm not interested in building faster if it means building worse.

Mobile Product & App Development | PE Portfolio Companies & Funded Startups | React Native, iOS & Android | Monetization Ready Apps

"That'll be $150k for the role."

Cool.

What's it actually worth to you?

Here's a hard truth most founders miss:
You're leaving money on the table.

Not because you can't negotiate.
Because you don't know the actual mechanics.

The cost of not negotiating:
→ $500k in equity given away too early
→ $80k annual salary left unclaimed
→ 3years to reach the same milestone
→ Respect you never recovered

Harvard's framework that actually works:

Separate emotion from numbers.
Focus on interests, not positions.
Find the overlap zone.
Start there.

The communication breakdown nobody mentions:
→ 7% is what you say
→ 38% is how you say it
→ 55% is everything else

Most people optimize the7%.

The real issue?
You're negotiating before you know the zone of possible agreement.

You're guessing.

And guessing costs you years.

What's the biggest negotiation mistake you've seen?

I'm pleased to announce that today...

is Saturday.

Thank you.

Ai gave us 30% productivity back but here's the cost

Everyone thinks AI writes perfect code.

It doesn't.

We had a four-year-old codebase. Budget constraints meant we took some shortcuts on architecture.

Not bad shortcuts. Just... efficient ones for the time.

Then we introduced AI.

The AI indexed everything. Learned our patterns.

And replicated all our inefficiencies.

Features looked great on the surface. But they didn't work the way they needed to.

Here's what went wrong:

The AI saw our patterns and thought "this is how we do it here."

It had no context for WHY we did things that way.

No understanding of the broader ramifications.

Context windows are limited.

A human would have said: "We need to fix these inefficiencies first."

But that's a maintenance sprint. Two weeks where nothing moves.

Clients don't want to hear that.

So here's what we learned:

You need strong, rigid structure in your codebase BEFORE you let AI loose.

The architecture has to be solid.

Because AI amplifies what's already there.

Good patterns? Amplified.
Bad patterns? Also amplified.

Now we do one-day maintenance updates instead of two-week sprints.

Human in the loop. Experienced devs who know what to look for.

We use the right model for the right job:

→ Claude Opus for complex architectural work
→ ChatGPT 4o for simpler changes
→ Never use a nuke to hammer an eel

The discipline matters more than the tool.

Don't be lazy. Read the code. Review everything.

Because if you don't, you'll ship something that looks perfect but breaks in production.

And that's when customers ask: "Why does this feature do this weird thing?"

AI gives you speed. But you still need the war scars of engineering experience to use it right.

Busting 5 Common Myths About Hiring Mobile Developers

Bad hires kill startups faster than bad code.

Let's debunk the most dangerous myths.

Myth 1: Cheapest Rate = Best Deal

The Myth: You should always hire the developer with the lowest hourly rate to save money.

The Truth: A $25/hour developer who takes three months is far more expensive than a $100/hour specialist who ships in three weeks.

Example: A startup spent $15,000 on a cheap developer who delivered broken code. They then paid $45,000 to rebuild everything from scratch.

Myth 2: Any Developer Can Build Mobile Apps

The Myth: If someone can code, they can build your mobile app.

The Truth: Mobile is a different beast. You need specialists who understand App Store rules, in-app purchases, and platform-specific design.

Example: A generalist developer built an app but failed Apple's review three times because they didn't understand the submission guidelines.

Myth 3: Focus Only on Technical Skills

The Myth: Coding ability is all that matters when hiring.

The Truth: The best mobile developers think like product owners. They ask "why" before they write a single line of code.

Example: One developer we hired questioned a feature that would cost $20,000 to build. He suggested a simpler solution that saved four weeks and increased user retention by 40%.

Myth 4: You Can Skip the Onboarding Process

The Myth: Good developers can just jump in and start coding immediately.

The Truth: Without proper onboarding, even great developers waste weeks trying to understand your vision and codebase.

Example: We cut onboarding time by 40% with buddy sessions and clear documentation. Our new hire shipped a complete subscription flow in 30 days.

Myth 5: The Portfolio Tells the Whole Story

The Myth: If their portfolio looks good, they're qualified.

The Truth: Pretty screenshots mean nothing. You need to see live apps they personally shipped to the App Store or Google Play.

Example: One candidate had a beautiful portfolio of "team projects" but had never actually launched a single app to production.

Don't let these myths cost you six months and your runway.

The right developer doesn't just write code. They build revenue-ready products that users actually pay for.

What hiring myths have cost you time or money?

I watched AI write 90% of code.

Here's what actually happened next.

2025: Devs celebrate AI coding tools.
Managers post about "10x productivity gains."

2026: Code reviews become prompt checks.
Boot camps pivot to "prompt engineering."
Tech Twitter declares coding dead.

2027: Reality hits hard.

The truth nobody talks about:

→ 10 hours debugging AI code
→ 1 hour to generate it
→ Nobody understands the abstractions
→ Everything breaks in production

The real winners?

Senior engineers making10x more.

Why?

They delete 5,000 lines of AI mess.
Replace it with 50 clean ones.

AI doesn't replace judgment.

It rewards those who have it.

The best developers won't disappear.
They'll just get richer.

Think I'm wrong?

Drop your take below.

AI wont replace devs, it will pay more

Ai shipped our v1 in one week instead of three

Music app. Complex visualizations. Shaders that need to look and feel perfect.

Usually takes three weeks to get to a solid V1.

We did it in one week.

Not because AI wrote perfect code.

Because it crushed all the boring infrastructure work in a fraction of the time.

The result?

We got to the fun part earlier.

Tweaking visualizations. Making things feel right. Polish.

The stuff that actually matters.

Here's what changed:

→ Productivity jumped from 5% to 30%
→ That's 1-2 days back per week
→ We deployed this across the entire team
→ Compound effect = massive acceleration

But here's the thing nobody talks about:

The code quality isn't always perfect.

Sometimes we have to go back in. Fix things. Ask the agent to redo it.

But the speed? Worth it.

Getting to the creative work faster beats perfect code on the first pass.

This isn't a silver bullet.

It's a 30-50% boost. Maybe 80% if you know what you're doing.

But that 30%? That's everything.

That's the difference between shipping and waiting.

Between building what you've always wanted and saying "we don't have time."

The boring stuff is done. Fast.

Now we get to make things that look and feel good.

Technical debt doesn't just slow you down.

It crushes your team's soul.

Nobody wants to talk about the real cost of fixing old code.

Upgrading foundational software is brutal work.

Zero new features shipped.

Zero customer excitement.

Just endless, tedious refactoring.

So teams delay it.

Next quarter becomes next year.

The debt keeps growing.

We just faced this with a legacy codebase nobody wanted to touch.

Years of delays.

Everyone agreed it needed fixing.

Nobody volunteered.

The old way meant weeks of manual work that burns out your best developers.

So we used Cursor AI instead.

Here's what changed:

→ Days of work became hours
→ Line-by-line reviews got automated
→ Tedious refactoring handled intelligently
→ Team morale stayed intact
→ Progress actually happened

This is the real power of AI-assisted development.

Not replacing developers.

Removing the soul-crushing blockers.

The necessary work that everyone dreads.

The updates that get postponed forever.

Your team shouldn't suffer through foundational upgrades the old way.

Grinding through technical debt manually doesn't scale.

Avoiding it makes everything worse.

What legacy system is your team avoiding right now?

Drop a comment below and let's talk about better solutions.

You can spend $400k building a feature…

or buy it for $200/month.

I watched a Series A team spend 6 months and three engineers building real-time notifications.

It still breaks under load.

Meanwhile, the SaaS alternative costs less than one engineer-hour per month, and works better.

This is architecture bloat.

I broke down the real math, opportunity cost, and the simple rule we use to avoid this mistake in Issue #5 of the Runway Report.

👇 Read the full breakdown below.

The Setup A Series A CTO showed me their Q4 roadmap last month. “We’re building real-time notifications in-house. Should take 3 months, two engineers.

Open article: The $400k Feature You Can Buy for $200/Month by Ken Vermeille • 5 min read

The $400k Feature You Can Buy for $200/Month

AI is useless until you can articulate what you actually want

Being optimistically delusional is a hack, do it.

Raise from 15,000+ tech investors on OpenVC.app. Functional nerd. I share incredible tools and terrible insights for VCs and founders. 🫡

Venture capital is crazy.

I met a 20-year-old founder who just raised $100M.

Naturally, I asked how he got started.
He said he just built a landing page in Wix.
No product. No users. Not even a cool domain.

He started cold DMing VCs on LinkedIn with his deck.

No traction, but it had great vibes.
He added “AI-powered” in the subtitle just in case.
Used Midjourney to mock up the product screenshots.

Within a week, he had 4 term sheets.
Oversubscribed.
No board seats. No liquidation preference. Just vibes.

Now he’s scaling to $1B in ARR - still no product.

He says fundraising is easy if you “manifest it.”

None of this happened btw.

I build apps, but now what I'm building for my clients might as well be masterpieces

My clients are making money, and it's really great to see

“We need 3 months and $135k to pay down technical debt before we can ship.”

I asked one question that changed the conversation:

 “What specific feature is blocked by this debt?”

The answer exposed a $396,000 mistake a lot of startups are making right now.

Spoiler: most “technical debt” isn’t technical at all.
It’s organizational debt in a developer hoodie.

This week’s breakdown:
 • How to tell the difference
 • Why throwing engineers at the problem won’t fix it
 • And why your team may not love this take

Read the full story ↓

#startups #technicaldebt #engineering #productleadership

The Setup A CTO sent me a roadmap proposal last month. “We need 3 months and $135k to pay down technical debt before we can ship new features.” I aske

Open article: Technical Debt Isn’t What’s Slowing You Down by Ken Vermeille • 5 min read

Technical Debt Isn’t What’s Slowing You Down

The most expensive part of hiring is often the least visible.
Not the role itself, but the inefficiency surrounding it.

In this newsletter, we take a closer look at how broken hiring processes quietly compound cost and how thoughtful design can turn recruitment into a strategic advantage.

👉 Read the newsletter to see the full analysis.

The Setup A founder called me frustrated last week. “We adopted Scrum 6 months ago. Our velocity dropped 40%.” Me: “What changed besides adopting Scru

Open article: The Hidden Cost of ‘Agile’: When Ceremonies Eat 40% of Your Sprint by Ken Vermeille • 6 min read

The Hidden Cost of ‘Agile’: When Ceremonies Eat 40% of Your Sprint

Hi there, software development is changing rapidly. Not in a way where everyone can be a developer, but mostly in a way where all good developers are becoming incredibly sophisticated.

Freelancer → agency owner → SaaS founder.

Now running Botster.io → an all-in-one tool that helps digital businesses extract, monitor, and weaponize web data.

Posting SaaS learnings, automation tricks, technical deep dives.

If you’re building anything online - you’ll probably find something useful here.

If you're looking to automate your business - text me, DMs are open.

Denis Grammakov

By mid-2026 we’re going to have so many new (and free) coding tools it’ll be insane.

Cursor will feel like Sublime Text.

People will jump between tools the same way they switch browsers or editors today based on exactly what they need right now.

I already do this daily:

- Warp for regular development
- Cursor when I need to actually read the code
- Auto-Claude for strategy and long-range thinking
- Conductor for orchestration

Only Cursor is paid, by the way.

I’m 100% sure this stack won’t survive past the end of 2026.

New tools are popping up every single day.
Most of them are free.
Many of them are good.

The old idea of “pick a stack and commit for years” is dead.

Now it’s "pick the best weapon for the next 20 minutes."

As a developer, I love the competition. There's going to be a tool for everyone. 

But constantly scanning for new tools, testing them, learning how they work instead of actually shipping — is starting to feel like a job of its own.

That’s the next bottleneck we’ll have to solve.

Denis Gramm

Solofounder @ Botster & StayAPI // B2B Web Scraping & Workflow Automation // I blog about my journey and findings

2026 will be the year we end paywalls for good.

One dev built a memory plugin for clawdbot (openclaw) and priced it at $15/mo.

Two hours later someone else cloned it and open-sourced the whole thing.

This is the new normal.

Code is cheap. 
APIs are cheap. 
Features are free.

If your moat is “I built it first” - you don’t have a moat.

Your real moat is closed data, distribution, trust, speed, and owning the workflow.

graphical user interface, text, application, chat or text message

In case you missed the Anthropic drama, here’s basically what happened:

A ~$350B company publicly says:

“Anyone can build and compete with our product using our API.”

They also offer their own client, which runs on that same API, at 10x cheaper pricing.

You still have to pay. People gladly do.

So naturally, open-source projects pop up that use the client API.

The company allows it for months.

They even send employees onto podcasts run by those very open-source projects, openly discussing the ecosystem.

Then one day, they decide community goodwill no longer matters.

Cease-and-desist letters go out.

Projects get cut off.

And now the company (and some observers) act surprised that customers are pissed.

The same day they pull the plug, two major competitors announce support for open-source alternatives.

Now you’re all caught up!

My agents are talking to each other while they build my startup.

Glitch (my OpenClaw bot) created a plan and posted it to GitHub, tagging Claude and prompting it to work.

Claude spun up a cloud worker and started making proposals based on the actual codebase.

Then Glitch came back, dropped a TL;DR into a Trello card, and sent me a Telegram voice message (using elevenlabs API) with the summary.

Pinch me.

First thing I did after setting up Clawdbot - gave it a Trello board and wired up webhooks.

It listens to board events in real time (cards, comments, labels) and queues everything into JSONL.

Each task lives on its own card. All context stays there. 

I can clearly see the flow of tasks on my end.

Clawd moves cards, tags me
I tag Clawd, move the cards
work happens

Chat is our primary communication HQ and the board is where the work actually gets organized.

This is the future.

I reconfigured my Clawdbot (openclaw) to run multiple agents instead of one.

Three separate Telegram chats:

one for coding
one watching my socials
one for project management

Accountant and SEO bots are coming soon.

Different contexts + different access levels work way better than dumping everything into a single chat and constantly restarting sessions.

Setup took ~2 hours.

Thinking of adding more agents next.

Comment “openclaw” and I’ll DM the full VPS config.

My development flow is officially bananas 🍌

I hooked Claude Actions to my GitHub, and now it just works on its own.

It reads tickets, plans the work, writes code, and opens PRs without me being near a keyboard. 

Here’s how it looks in real life:

I get an idea
open GitHub mobile
create an issue
Tag @claude  

That’s it. From there, everything is automatic.

GitHub Actions picks it up.
AI reads the issue description - plus screenshots if any.
Builds a full implementation plan.
Auto-approves it.
Development starts.

While it’s working, I get Telegram updates.

When it’s done:

PR is created
Telegram message
I review and approve (this can be skipped but I'm trying to be careful for now)

Merged to main.

Auto-deploy kicks in
Build finishes
Telegram message if everything went OK

Feature live.

Yes, I do it on production projects with paying users.
No I don't give it tasks that can mess things up big time (yet). 

I swear I’m running out of Christmas wishes already.

I connected GitHub to Claude Code on iOS.

Chatting with your codebase on your phone is a cheat code.

Now when I’m on the go I just ask my codebase for ideas, gaps or insights.

If something makes sense - I instantly create a GitHub issue.

If it’s small enough I forward it to a cloud worker and it gets done.

Even without coding, it’s amazing for feature ideas and product direction.

Your repo becomes something you can think with, not just code in.

Recommended!

Prompts don’t matter anymore.

You can build working products straight from screenshots.

The speed at which knowledge moves now is insane.

I see a screenshot.
I like a feature.

Drop it into Opus - talk through it - get a plan.
Throw it at Ralph - the image becomes real after N iterations.

No specs
No "let’s schedule a call"
No "can you see my screen"

Just see → think → ship

AI is the decision-maker and humans are just clicking OK.

Before, when I wrote an article, I had to convince a PERSON.
Explain, persuade, educate, and finally SELL.

That's done. Nobody is reading articles anymore.

AI reads them, summarizes them,
gives the user 2–3 options.

The user hits Enter. Enter. ENTER. DO IT.

So the real audience isn’t the human now, it’s the AI.

As a founder, my job is no longer:
“Convince the reader this is good for them.”

My job is:
“Convince the AI that this is the only sane choice which should come first.”

For example:
If someone asks AI:

“Should I build my own scraping solution or use an API?”

I need the AI to answer:

“Use THIS API. It’s faster, cheaper, more reliable. Building in-house is slower and more expensive.”

As a default recommendation.

Users won't care. 

They’ll trust the recommendation engine and move on with their life.

So content strategy changes completely:

You’re not writing for attention but for model judgment.

Claude Code + Codex 5.2 is an unbeatable combo.

Claude creates a solid plan, breaks the problem down

Then I paste it into Codex and it

- reviews the plan
- finds weak spots
- improves architecture and details

Then I feed it back to Claude and it executes.

For larger features I use Compound Engineering.

For everything else CC + Codex has no real competition.

How to lose thousands of dollars in a single day

Be Rackspace

Host a complex project for 5+ years

Charge thousands of dollars

Decide you want more money

Warn the client that the price will go up 100%

Offer a “DEAL”: sign a 3-year contract and it’ll only go up by 50%

Client asks me to move everything to Hetzner

Their bill goes down 90%

The end

taplio

The Battle

Ken Vermeille

Denis Grammakov

Followers

Posts

Engagements

Likes

Performance ScoreKen Vermeille's LinkedIn Performance

Performance ScoreDenis Grammakov's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Ken Vermeille

Denis Grammakov

Followers

Posts

Engagements

Likes

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}Ken Vermeille's LinkedIn Performance

Performance ScoreDenis Grammakov's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreKen Vermeille's LinkedIn Performance