Dharmesh Shah

Founder and CTO at HubSpot. Helping millions grow better.

Career startup guy.

Author of the book &quot;Inbound Marketing&quot; (http://InboundBook.com)

Specialties: Entrepreneurship, Software Development, internet marketing,SEO

Founder and CTO

Founder of HubSpot, a venture-backed software startup.  HubSpot offers the industry's first inbound marketing system for small businesses.  The software is available as a hosted service (SaaS).

This might be the SCARIEST and/or CRAZIEST thing I've done in years.

But, here we go...

Even before PLG (Product Led Growth) there is usually Founder Led Growth.

I've long believed that *everyone* in an early-stage startup should be selling the product and helping customers get value. Even if you've never done sales. *Especially* if you've never done sales.  

But, there's no reason to limit this to just the early days of a startup.

So, in the spirit of curiosity, learning and love, I'm trying something new.

I'm going to jump back into "early-stage" mode and help sell HubSpot. Not just sell the vision of HubSpot (which I do all the time), but sell the actual product/platform that we provide to customers.

HubSpot is 20 years old and has 280,000+ customers. Why would I do this? 

Simple, because:  I LOVE HUBSPOT! 

And, I want to share my love with as many other founders and founding teams as I can. 

Also because I'm competitive and love a challenge.  

I'm going to track it all and see how many people I can get to sign-up for HubSpot Free or Starter edition. 

I've got a ton of ideas I want to try that I think will help drive sign-ups. 

As part of this effort, I'll be creating a series of "founder to founder" videos. Some ideas I have:

1) A "speed-run" of signing up for HubSpot and getting your first contact in there and starting to use it.

2) A demo/screencast of how I *personally* use HubSpot for Agent.ai (a HubSpot Next project).  It has 2.5M users and is growing. So, it's a *real* startup with real data. I get to also then show-off how we're using HubSpot with AgentAI to automate some things (like monitoring for VIP signups). 

3) WTF is an ACP (Agentic Customer Platform) and does a startup really need one?

4) Why can't I juse vibe code my own CRM?

5) Why wouldn't I just use one of the super-cool AI-native CRM products? 

Or...any questions that you folks have.

Phew!  I'm nervous...but excited.  

If you're already a HubSpot customer, THANK YOU! 

If not, then give me a chance to convince you.  :)

Visit:  GOHUBSPOT .ai

And in the meantime, please leave your comments, feedback, questions and pushbacks for upcoming videos.

Off to the races!

Thanks for your support.

#ILoveHubSpot #GoHubSpot

Woo hoo!  Thrilled to share a sneak peek of HubCode -- The Vibe Coding Tool for HubSpot.

I talked about this a few weeks ago (which feels like forever). 

Been cranking on it and am ready to show you the actual app now.  Link to video in the comments. Or, just type video.hubcode .com in your browser (it takes you to YouTube).

Some notes:

1) This requires no terminal -- runs completely in the cloud

2) You can vibe code both app cards and full apps that live inside the HubSpot web UI.

3) Generated apps can access data within HubSpot (of course).

4) SUPER EXCITING: Apps can reuse any of the public agents in agent .ai (there are 2,000+ of them). This makes your apps super-powerful because you can access all sorts of data and services. And, you can build your own agents and access them too.

This gives you access to all the major frontier models, a bunch of really useful data sources and a ton of other capability. Go to agent .ai to see what's available.

The demo shows pulling YouTube videos related to a company right into the app card.

5) You can paste in screenshots or other visuals as part of your vibe coding process (if you haven't done this before, it will feel magical). 

The HubCode app runs on the agent .ai platform and is in private beta (because this one's going to take some testing and iteration give the power and flexibility).  To apply, just go to hubcode .com.

This is an early "proof of concept", but will be iterating on it maniacally as I get user feedback.

Let me know what you think.  All thoughts, ideas, feedback and wishlist items appreciated.

FEEDBACK IS A GIFT.  Thank you!

Thank you for your support.

I am pioneering the Future of Reskilling, at global scale, with urgency, and with a clear blueprint to prepare humanity for the AI age.

Across Fortune 500 boardrooms and refugee camps alike, the need is the same: talent that is AI-literate, outcomes-driven, and future-ready. That’s where I focus my life’s work.

With two decades of leadership at Infosys across the Americas and Europe, I’ve led transformation at the intersection of business, technology, and human capital. But in 2019, I turned that experience toward a more ambitious goal: to reskill and employ 1 million underserved individuals, especially youth in refugee camps and overlooked communities, using AI not just as a tool, but as a force multiplier for economic mobility.

Today, through StepUp.One and Human-Edge.ai, I’m building a new kind of workforce infrastructure, one that is digital, distributed, and deeply human. We partner with global enterprises, development agencies, and impact-focused leaders to deliver results, not just programs.

At the heart of our model is a 3-part Reskilling Formula designed for real-world ROI:

Master the Topic – This is where Today’s Learners Stop when they get their certificates.

Master the Skill – Apply the knowledge to deliver outputs. Use AI to amplify your results.

Master the Outcome – Drive Outcomes from Outputs. When they don’t come, Change your strategy, structure and execution.

This framework works equally well for a VP in New York, a jobseeker in Nairobi, or a student in São Paulo. It bridges the global reskilling gap not with hype, but with orchestration—pairing advanced AI agents with trained human operators to deliver performance at scale.

Our conviction is simple but radical:

Human + AI = Outcomes

But reskilling is only the starting point.

We go further by embedding this trained talent into AI-powered delivery pods that support everything from enterprise sales and marketing to fundraising, recruiting, and digital operations.
Whether it’s helping a Silicon Valley startup scale faster, or enabling a Fortune 500 firm to enter a new market with agility, our model proves this: underserved populations, when equipped with AI and structured support, can outperform traditional teams at a fraction of the cost.

If your mandate is to lead transformation—across markets, models, or mindsets—let’s talk. Because the future of work isn’t just about automation. It’s about inclusion, leverage, and creating a globally distributed workforce that wins.

Let’s build that future—together.

Mohamed Anis

After 200+ fundraising campaigns, we've turned the art of raising capital into a repeatable science.

Here's the playbook that transformed fundraising from art to science →

View newsletter: Create Your Human Edge with AI

Ever wondered why some founders raise millions with a napkin pitch while others struggle despite having traction, revenue, and a polished deck? After 

Open article: Helping First-Time Founders Raise Capital by Mohamed Anis • 3 min read

Helping First-Time Founders Raise Capital

Here's the case for why Claude Code + COBOL could unwind a century of IBM dominance and why this time it's structural, not cyclical.

IBM's moat was never technical. It was cognitive.

800 billion lines of COBOL sit in active production. $3 trillion in daily commerce flows through it. 95% of ATM transactions run on it. IBM didn't protect that kingdom with patents. It protected it with complexity, deliberately accumulated, institutionally embedded, humanly irreplaceable.

The playbook worked for 65 years because COBOL expertise takes 20+ years to develop meaningfully. The average COBOL developer is now 55+. That's not a talent pool. That's a ticking liability.

IBM Consulting built a $20B+ annual business on a simple arbitrage: enterprises couldn't touch the systems themselves, so IBM charged them to maintain, extend, and occasionally modernise them, on IBM's timeline, at 

IBM's rates, with IBM-certified humans.

AI doesn't just speed that up. It eliminates the arbitrage entirely.

When a model can read, reason about, and translate COBOL at a human-expert level, the knowledge scarcity that created IBM's pricing power disappears. Not gradually. Suddenly.

What previously required 3-5 year multi-million-pound programmes could be compressed to months. What required IBMers with 20 years of mainframe scar tissue can now be scaffolded by a junior engineer with Claude Code and good judgment.

Three pillars held IBM's moat:

1 Proprietary tooling - still relevant, but eroding as AI-native tools match output quality

2 Certified expertise scarcity - gone when any competent engineer can query the model

3 Enterprise risk aversion - the last standing wall, but Tier-1 banks are already running pilots

You're living through pillar three cracking in real time.

The real IBM risk isn't the Z17. It's the consulting P&L.

The Z platform's 40% growth is real, and IBM Z17, supporting Java and modern workloads, is a smart hedge. But hardware is not where the margin lives. IBM Consulting is. And consulting revenue requires duration. long programmes, high headcount, multi-year contracts.

When AI compresses a 5-year engagement into 8 months, IBM doesn't get 5 years' worth of fees on a smaller deal. It doesn't get the deal at all.

This is the Eastman Kodak Company moment - not because the product is bad, but because the problem it solves is shrinking.

Jasper Tech Ai, Gamma, Cursor, yes, they'll face the same gravity. But they were born in the AI era. IBM built its entire identity on a problem that required human scarcity to remain monetisable.

The 13% drop isn't a panic. It's the market slowly understanding that IBM's core value proposition-"we are the only ones who can safely touch your most critical systems" just had its first genuinely credible challenger.

That's not a dip. 

That's a re-rating of what IBM is worth in a world where the moat can be drained.

Seven years ago, nobody was hiring from Kakuma or Dadaab.

Today, women from these camps run client campaigns across continents. 

They use AI tools that most professionals haven't yet figured out. 

They Learn
They Earn.
They Train others.

No pity. No handouts. Just work.

This International Women's Day, I'm not writing about them. 
I'm letting them speak.

Voices from Kakuma and Dadaab:

"My first client didn't know I was in a refugee camp. 
They just knew their leads doubled."

"I learned AI tools before my internet was stable enough to run them."

"My mother carried me across a border. 
I'm carrying her back with a salary."

"They told me a refugee can't do fundraising for a Silicon Valley startup. 
The startup disagrees."

"I trained 4 women last quarter. 
Two of them now earn more than I did when I started."

"The camp gave me shelter. 
StepUp.One gave me a job title."

"My daughter saw me on a call last week.
She asked if I was a boss. I said yes."

"I used to wait for aid. 
Now clients wait for my deliverables."

"Nobody asks where I'm from anymore.
They ask when I'm available."

"I don't want to be an inspirational story.
I want to be on the next project."

These women don't need a day. 
They need more clients.

If your company needs execution in marketing, sales, or fundraising, these are the people who will deliver it.

Finally, none of this would have been possible without the support of my wife and my daughter, who not only inspired me to do this but also continue to support me through the toughest times of our lives.

Happy International Women's Day.

#IWD2026 #WomenInAI #RefugeeEmployment #Kakuma #Dadaab #StepUpOne

Every major company now has access to GPT-5, Claude, and Gemini.

Same models. Same APIs. Same pricing tiers.

So why are results so wildly different?

I've been thinking about this since reading a recent HBR piece by Rohan Murty and Ravi Kumar S. 

They studied 50+ large enterprises and found something that should concern every leader investing in AI right now.

Companies using the same tools, performing the same functions, in the same industries, produced consistently different outcomes. The variation wasn't explained by which model they picked. It was explained by context.

Not "context window." Organizational context. The patterns of judgment, coordination, and trade-offs that shape how work actually gets done inside a specific company.

That's the variable nobody's optimizing for. And I think it's the only one that matters.

More on this tomorrow.

The Question Nobody's Asking

Our highest earner never speaks in our internal team meetings.
Not once in months.

But in 3 years, he has never missed a single weekly client report. 
Not one week. 
Not one client. 
Not one excuse.

Meanwhile, the loudest performers? 
Winning clients left and right. 
Losing them just as fast.

So we did the math on our own team:

Winning a new client = 10% weightage 
Retaining a client = 70% weightage

Retention is 7x more valuable than acquisition.

Everyone knows this intellectually. 
Almost nobody builds their scoreboard around it.

We were celebrating wins. 
We should have been celebrating stays.

The person who never speaks in your meetings might be your most valuable player. 

What does your scoreboard actually measure?

#HumanPlusAI #GTM #ClientRetention #StartupLessons

Dear Team Anthropic,

Please consider releasing simple videos like this to make us get the gist of what you are trying to convey in the 20-page report Or Did i just miss it?

StepUp.One builds custom AI that thinks like your business trained on your data, speaking your voice, validated by humans who've scaled companies before. 

They don't sell software; they sell results: revenue, fundraising, market position - measured, optimized, de-risked. 

AI does the work, experts own the outcome.

Yann LeCun just raised $1.03B for AMI LABS.

His thesis: LLMs learn from language, not reality. 

That's a fundamental limitation. 

To get real intelligence, you need world models - AI that understands how the physical world works.

He's right.

But there's a second limitation: nobody is funding it.

LLMs don't just fail to understand the world. 

They fail to understand you. 

Your context. 

Your job. 

The domain you're operating in. 

The unwritten rules. 

The relationships. 

The trade-offs.

Yann LeCun is building models that understand the world.

We're building infrastructure that understands the human.

$1B just went to world models.

~$0 has gone to context infrastructure.

Both are correct that the model alone is not enough.

Both are building layers that sit around the model, not a better model.

The complete stack needs both.

But right now, one side of the equation is funded.

The other isn't.

The game isn't keywords anymore - it's context, structure, and trust signals that LLMs can parse and cite. 

If your content doesn't answer questions clearly, use credible sources, and follow a logical hierarchy, you're invisible to the AI layer where 94% of B2B buyers now start their research. 

Optimize for machines that read like humans: be direct, be cited, be structure dor be skipped entirely.

The best talent isn't waiting in traditional labor markets. 

It's globally distributed, underutilized, and ready to outperform.

Most organizations treat AI adoption and workforce transformation as separate problems.

They're not.

When you build AI that thinks like your business, is trained on your data, and speaks your voice.

The transformation isn't about replacing people with AI. 

It's about amplifying human capability at scale.

AI does the work. 

Human experts who've actually scaled companies own the outcome. 

That's the difference between hype and results.

The HBR article tells a story I can't stop thinking about.

Two large B2B companies. 
Both sell complex, multi-year technology services. 
They compete for the same enterprise customers. 
Same sales stages. 
Same forecasting cadence. 
Same CRM setup.

Look at their systems, and they're identical.
But in practice, they're completely different.

Company A serves regulated industries. 
Their sellers pause deal advancement while delivery leaders review feasibility in parallel. Risk and legal get looped in early, even when terms are still fluid.

Company B serves digitally driven enterprises. 
Speed wins. Salespeople pull delivery and solution leaders into exploratory conversations before the scope is finalized. 
Leaders intervene early to shape direction, not validate certainty.

Here's what struck me: neither company's CRM captures any of this. 

The system records stage changes. 
It doesn't record how feasibility was debated. 
Which risks were weighed? 
Why was escalation deferred?

Those decisions live in emails, chats, spreadsheets, and conversations. And they largely disappear once the deal moves forward.

Yet this is exactly what makes each company successful in its market. 

The CRM shows what happened. 

It never shows how or why.

Outputs vs Outcomes is already a massive debate, but the bigger problem now is cutting through the AI noise.

I am sure very soon there will be an AI that will be reading all of my social media streams and curating the ones that i must see or respond.

Cursor charges $200/month for Claude Code.

It costs them up to $5,000 in computing to deliver it.

Read that again. 

For every $1 you pay, someone is burning $25 behind the scenes.

Last year it was $2,000. This year $5,000. The cost is going UP, not down.

Where's the difference coming from? Venture capital.

VCs are subsidising your AI productivity right now. 

Every smart autocomplete. 

Every code generation. 

Every agent that writes your tests. 

Someone is writing a cheque so you can have it for cheap.

We've seen this movie before.

Uber did it with rides. 
WeWork did it with desks. 
DoorDash did it with deliveries.

Subsidise like crazy. Capture the market. Then reprice.

But here's what's different this time.

Cursor doesn't control the models. 
They're reselling Anthropic and OpenAI's compute at a loss. 
They can't subsidise their own users the way Anthropic can subsidise Claude Code directly.

That's a brutal position to be in. 
Your entire margin depends on someone else's pricing decision.

The platform eventually competes with you. Or reprices you out.

So what should builders do right now?

Use everything. Build fast. 

This window of subsidised AI is real and it won't last.

But don't confuse subsidised pricing with sustainable economics.

The companies that win long term will be the ones who turn this cheap compute into durable client relationships and measurable outcomes before the subsidy disappears.

Cheap AI is temporary. 
The outcomes you build with it don't have to be.

#AI #Startups #Founders #VentureCapital #AIStrategy #ClaudeCode

This is a massive validation of the entire Human-Edge.ai thesis, and I don't think it's obvious at first glance why. 

Here's what makes this a massive deal for us specifically:

1. Karpathy validated the three-layer thesis.

His system has exactly three layers: the frame (program.md), the execution engine (the agent + model), and the feedback loop (val_bpb → keep/discard). Remove any one, and the system breaks. That's our equation: Domain Operating Frame + Model + Feedback Loop = Outcomes. He just proved it works in ML research. We're building it for business domains.

2. Garry Tan's line is our entire company thesis.

"You can't just ask the agent to self-improve. You have to design the arena and give it a push based on your knowledge of how it all works." That IS the Domain Operating Frame. The arena design is the product. 

For Karpathy, the arena is a 5-minute training budget with a single metric. For Human-Edge, the arena is the fundraising operating frame with multi-dimensional outcome tracking.

3. "The best program.md wins" = "The best context infrastructure wins."

Garry Tan's closing argument: "A year from now, the companies winning won't be the ones with the most engineers or the most compute. They'll be the ones whose agents never stopped running. The best program.md wins." The best Domain Operating Frame wins. The best Human Graph + Job Graph wins. The model is a commodity. The arena is the moat. That's our entire investment thesis in one sentence from someone else's mouth.

4. The progression timeline maps perfectly to our Generations.

Karpathy's own timeline: write code (2024) → tell AI to write code/vibe coding (Feb 2025) → tell AI how to think about writing code / agentic engineering (Feb 2026) → write a Markdown file and let AI iterate autonomously / autoresearch (March 2026). 

That is Generation 1 → Generation 2 → Generation 3 in our paper. 

Raw prompting → prompt engineering → context-driven autonomous execution.

5. The critical difference — and our opportunity.

Karpathy's arena has one clean metric: val_bpb. 
That makes the feedback loop trivially clear. 

Business domains don't have that. Fundraising success is multi-dimensional: speed, valuation, dilution, strategic fit, founder time cost, and relationship preservation. 

This is precisely why the Domain Operating Frame must be engineered, not just written in a Markdown file. Karpathy can get away with a flat .md because ML research has a clean loss function. 

Business doesn't. The operating frame for fundraising must encode the ontology, the workflow state, the constraints, the actor archetypes, the norms, AND the multi-dimensional objective function.

Anthropic just gave Claude a whiteboard.

For free. 

On every plan.

Let's talk about who should be worried.

Claude now builds interactive charts, diagrams, flowcharts, and visualizations inside the conversation. No extra tool. No subscription. No learning curve. You ask a question. It draws the answer.

Here's the body count:

🔹 Miro ($17.5B valuation, $500M+ ARR) Sells collaborative whiteboards and diagrams at $10-20/user/month. Claude now generates diagrams on demand. For free. Miro's entire value proposition is "visualise your thinking." Claude just made that a side effect of thinking.

🔹 Lucidchart ($210M revenue, $3B valuation) Built a business on flowcharts, org charts, and process maps. Claude builds those mid-conversation now. No drag and drop. No template library. Just describe what you need.

🔹 Canva ($42B valuation, $4B ARR) The charting and infographic side of Canva just got a free competitor baked into every Claude conversation. Canva still wins on design-heavy assets. But for quick data visualisation? Why would you leave the chat?

🔹 Whimsical, Excalidraw, draw.io These tools live on simplicity. "Just draw something fast." Claude now does that faster. In the same window where you're already working.

🔹 Google Jamboard Already dead. Google shut it down in 2024 and pointed users to Miro and Lucidspark. Now both of those are under pressure from the same AI that Google couldn't build fast enough.

🔹 Basic BI dashboards If you're using Tableau or Power BI just to make a quick chart for a meeting, Claude does that in 10 seconds now. The enterprise BI platforms won't die. But the casual use case? Gone.

The pattern is always the same.

Standalone tools get absorbed into platforms. Features that used to justify a $20/month subscription become a free capability inside something bigger.
Miro didn't kill whiteboards. It digitised them. Now Claude is absorbing the digital version into the conversation itself.

The companies that survive this will be the ones who own workflows that go deeper than "make me a chart." Collaboration, version control, team permissions, enterprise compliance.

But for the 80% of users who just need a quick visual to make a decision?
They just lost every reason to open a separate tool.

This is how bundling works. Not with an announcement. With a feature that makes you forget you ever needed the other product.

#AI #Claude #Anthropic #Miro #Canva #Lucidchart #Startups #Technology

https://lnkd.in/eVrAjMgq

Open video: Claude can now show you on youtu.be

For nearly six years, we told the market that we do three things.

Help startups, scale-ups, and enterprises raise capital, scale GTM, and hunt talent.

All true.
We did all of it.
We were good at it.

But our growth was slow. 
Our pipeline was unpredictable. 
Our messaging confused people.
We sounded like everyone else.

Six months ago, we made a decision that changed everything.
We niched down.

Not in what we deliver. In what we say.
We dropped GTM and Talent from our marketing completely. 
Both still account for 50% of our revenue. 
But we stopped talking about them.

Instead, we went with one line:

We help first-time founders raise their pre-seed from investors who have already invested in their niche.

One audience. 
One problem. 
One outcome.
Here's what happened.

Our CRM went from a slow drip to over 1,300 live opportunities.
In six months. I attribute this single-handedly to niching down.

The niche is the door. 
Everything else is the house.

Here's what I got wrong for six years.

I thought saying less meant being less.
I thought if I didn't mention all three services, I'd lose two-thirds of my market.

The opposite happened.

By saying one thing clearly, we attracted more of everything.

Your niche isn't about shrinking what you do.
It's about sharpening what you say.

What's the one line that describes your company right now? 
Could a stranger repeat it?

#Startups #Fundraising #NichingDown #PreSeed #FounderLife

Anthropic published a labor market report this week. 

It's the most rigorous measurement of AI's impact on jobs I've seen. 

Real usage data from Claude mapped against 800 occupations. 
Careful methodology. 
Honest findings.

Computer programmers: 74.5% of their tasks are now covered by AI. 

Customer service reps: 70.1%. 
Data entry: 67.1%.

No systematic rise in unemployment yet. 

But hiring of workers aged 22 to 25 has dropped 14% in the most exposed occupations.

The report is excellent. 
And I think it's asking the wrong question.

The entire framework measures tasks. 

Which tasks can AI perform? 
What percentage of each job is "covered"? 

Track whether unemployment rises for the most-covered workers.

But a job is not a bundle of tasks. 

A job is a bundle of tasks wrapped in context.

A programmer who writes code is performing a task. 

A programmer who knows why a specific architecture decision was made three years ago, which parts of the codebase are fragile, and why a feature request will take longer than expected because of undocumented technical debt. That's not a task. That's context. 

It's why that person is worth $180K a year instead of $0.003 per API call.

That 14% hiring drop? 
Everyone is reading it as a labor market signal. 
I read it as a context crisis.

When a 23-year-old joins a firm as a junior analyst, their year-one tasks are exactly what AI does well. Pull data. Build models. Format reports. Those tasks show up in Anthropic's usage data.

By year three, that person isn't valuable because they build models. They're valuable because they carry context that no system ever captured.

Stop hiring those 23-year-olds because AI handles their year-one tasks, and you're not just saving salary. You're cutting the pipeline through which your organization builds and transmits institutional knowledge.

You'll feel it in five years when you have tools that execute tasks flawlessly and nobody left who understands why certain decisions should be made in the first place.

The report also found something that inverts the usual automation story. The most exposed workers aren't the least skilled. They earn 47% more than the unexposed group. They're nearly four times more likely to hold a graduate degree.

That wage premium isn't paying for task execution. It's paying for context. Relationships. Client history. Domain expertise that lives in someone's head, not in a system.

AI automates tasks. It does not automate context. And right now, almost nobody is building infrastructure to capture, structure, and preserve the context that makes experienced workers valuable.

That's what we're building at Human-Edge.AI. Context infrastructure starting from the individual human. Voice, identity, connections, ideas, professional trajectory.

For most knowledge-work roles, the green layer is bigger than the blue.
It's unmeasured. It's unvalued. And it's eroding.

That's the problem worth solving.

They're Measuring the Wrong Thing

Google Isn't Building a Figma Killer. It's Doing Something Worse.

Figma went public in July 2025. Market cap hit $56 billion.

Eight months later? $12.6 billion.

That's 77% gone. While revenue grew 40%.

This week, Google launched the latest version of Stitch. An AI-native design canvas. You describe what you want. It builds full UI screens. From text, voice, screenshots, code. All at once. Free.

You don't start with a wireframe. You start with "I need a landing page that feels calm and minimal." Stitch gives you multiple directions. Instantly.
Then you talk to it. "Give me three menu layouts." "Dark palette." "Stitch these into a clickable prototype."

What took three days now takes three minutes.

Figma stock dropped 12% in two days after the announcement.

Here's what Google actually shipped:

Infinite canvas that ingests images, text, and code as context simultaneously
An AI agent that tracks your whole project, critiques designs, and generates variations

Instant interactive prototypes from static screens

DESIGN.md that pushes your design tokens directly into dev tools

Deep wiring into Gemini, Antigravity, Claude Code, and Cursor via MCP
Google didn't build a better design tool. They turned the design file into a by-product of talking to their AI.

And they can give it away for free. Because Google doesn't need design revenue. They need you using Gemini tokens and Google Cloud.

Figma needs per-seat pricing to survive. That's not a fair fight.

Same pressure is building on Framer. $2 billion private valuation from August 2025. Strong product. But if Stitch makes it trivial to go from prompt to production UI, the standalone design-to-site category gets very hard to defend.

The pattern repeating across SaaS in 2026:

AI Platform + Distribution + Free Pricing = Category Compression

$285 billion wiped from software stocks in February. Jefferies called it the "SaaSpocalypse." And Google just made professional UI design a free feature inside its AI stack.

Now. I use Figma. I've used it for years. I still think it's one of the best products ever built for collaborative design. The community, the plugins, the muscle memory, the handoff workflows. None of that disappears overnight.

So I'm genuinely asking the designers here:

Where is Figma's moat now? Is it the network? The design systems? The enterprise lock-in? Or is it just habit?

Because the data says the market doesn't believe the category anymore. And I want someone to tell me why the market is wrong.

Argue for Figma. I want to hear it.

#Figma #Google #VibeDesign #AI #SaaS #ProductDesign #Stitch

OpenAI just published a study that most people will scroll past.
But it tells you everything about where AI actually stands today.

They tested whether their smartest reasoning models could control their own thinking process.

Not their answers. Their thinking.

Could they hide certain words from their chain of thought? Could they follow simple formatting rules while solving problems?

The scores? Between 0.1% and 15.4%.

Let that sit for a second.

The most advanced AI models on the planet cannot reliably control how they think.

And here's the part that stuck with me.

The models would NOTICE they were breaking the rules. They'd say "wait, I wasn't supposed to use that word." And then they'd use it again. And again. And again.

One model panicked mid-reasoning, realized it violated its own instructions, tried to course correct, and failed four more times in a row.

Now OpenAI says this is actually good news.

Why? Because if AI can't control its own thinking, it also can't hide dangerous reasoning from safety monitors.

Fair point.

But I see a bigger lesson here.

If the best AI models on the planet can't even control their internal reasoning process, why are we pretending they can run our businesses, close our deals, deliver our client outcomes?

The Illusion of Algorithmic Self-Control

Your CRM stores that a deal has moved to "Negotiation."

Your ERP stores that a purchase order was approved.

Your ticketing system stores that a case was resolved.

None of these systems stores the pattern of coordination, deliberation, and trade-offs that actually produced those outcomes.

This isn't a documentation problem. 

You can't fix it by adding a notes field.

It's a structural limitation. 

These systems were designed for record-keeping, not for reasoning. 

They capture what an organization decides, but not how it decides.

And that "how" is everything. 

It's the accumulated judgment of hundreds of people making thousands of small decisions over years. 

It's what makes your company yours, not a copy of your competitor's with the same tech stack.

I keep coming back to this gap. 

We've spent decades digitizing outcomes. 

We've barely started digitizing understanding.

taplio

The Battle

Dharmesh Shah

Mohamed Anis

Followers

Posts

Engagements

Likes

Performance ScoreDharmesh Shah's LinkedIn Performance

Performance ScoreMohamed Anis's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Dharmesh Shah

Mohamed Anis

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);}Dharmesh Shah's LinkedIn Performance

Performance ScoreMohamed Anis's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreDharmesh Shah's LinkedIn Performance