UnoMiru

27/08/2026

BECO didn't just launch an ad. It started a conversation.

BECO went straight after two of India's biggest home-care brands — Surf Excel and Vim.
It named the competitors.

Used their famous brand lines.
Put ingredient claims front and centre.
And asked consumers to look at what's inside.
Then HUL took the campaign to court.
And suddenly, the ad wasn't just an ad anymore — it was the story.

That's what makes this campaign so interesting.
Old marketing:
“Our product is better.”
The new challenger playbook:
Name the competitor → challenge the category → create debate → get people talking → let the conversation grow.

BECO didn't simply compete for attention.
It created a reason for everyone else to talk about it.

Is this where FMCG marketing is headed? 👀

08/08/2026

NVIDIA is betting that the next AI advantage may not come from making models bigger.

Its new Nemotron-Labs-3-Puzzle-75B-A9B takes Nemotron-3-Super from roughly 120.7B total parameters to 75.3B, while reducing active parameters from 12.8B to 9.3B.

But the interesting part is what happens after the compression.
NVIDIA reports around 2× higher server throughput on a single 8×B200 setup under matched user throughput conditions.
On a single H100, its 1M-token context test increased sustainable concurrency from 1 request to 8.

That matters because AI is moving from occasional prompts to agents, long-context workloads and systems that need to run continuously.

A smarter model is useful.
A smarter model that is cheaper and easier to serve at scale could be much more important.

The bigger question is:
Are we entering a phase where AI efficiency matters as much as AI intelligence?

What do you think?

30/06/2026

Most celebrity brands start with a famous face.

One8 tried to build something that could stand on more than just one.

What stood out wasn't only the product—it was how the brand invited people into the journey.

Design.

Craftsmanship.

Manufacturing.

The launch became more than a product reveal; it became a story about how the product came to life.

That changes the conversation.

Instead of asking,
"Which celebrity owns this?"

People begin asking,
"How was this actually made?"

In a world where polished ads, influencer campaigns, and AI-generated content are everywhere, showing the process can become a competitive advantage.

People don't just buy products.

They buy confidence.

They buy craftsmanship.

They buy proof.

What builds more trust for you?

A celebrity endorsement or seeing how the product is made? 👇

20/06/2026

2,000,000 tokens.

That is Google's new AI context window.

To put that in human terms:
The entire Harry Potter series is about 1.1 million words.
Gemini 3.1 Ultra holds twice that in a single conversation.

You could feed it every email you've sent this year.
Every document in your company drive.
Every transcript from every meeting since January.

And ask it anything across all of it.
At the same time.

But the context window isn't even the most interesting part.

Most multimodal AI does this:
Audio → transcribes to text → processes text → answers.

That chain introduces latency. It loses nuance. It misses tone.

Gemini 3.1 Ultra doesn't do that.

It processes text, image, audio, and video natively — simultaneously.

It doesn't read about the sound. It hears it.
It doesn't read a description of the video. It watches it.
It doesn't get a summary of the image. It sees it.

All at once.

Then there's the sandboxed Code Ex*****on tool.

Ask Gemini 3.1 Ultra to write a script.
It writes it.
Runs it in a sandbox mid-conversation.
Shows you the output.
Catches the errors.
Fixes them.
All without leaving the chat window.

For developers: the write → test → debug cycle just became one conversation.
For non-developers: describe what you want, and an AI builds and verifies it for you.

Most model releases are just bigger versions of the same architecture.

Gemini 3.1 Ultra is architecturally different in three separate ways.

That doesn't happen often.

What would you use 2 million tokens of context for first?

Drop it below. 👇

19/06/2026

Indian Startups. Week of June 9–16, 2026.

The headline wasn't a funding round this week.
It was an IPO filing.

Zepto filed its updated DRHP on June 9.
Fresh issue: ₹8,010 crore.
FY26 revenue: ₹22,623 crore doubled in one year.
Founded: 2021.
Listing age: approximately 5 years old.

One of the fastest startup-to-IPO timelines in Indian history.

But strip out Zepto and the deeper story of this week is the shift in where capital is going.

SolarSquare: ₹635 crore Series C.
Investors: IFC, Siemens Financial Services, Fullerton Carbon Action Fund.
Not a consumer app. Rooftop solar infrastructure.

BazaarNow: ₹72 crore from Peak XV.
Founded January 2026. Five months old.
1,800 orders per day per store in pilot city.
Built by ex-Zepto executives for Tier 2 and Tier 3 India.

Svaraa: $9.7 million from Accel and NKSquared.
Autonomous ships and underwater defence platforms.
Founded by a naval architect and a Rear Admiral.

ContraVault AI: ₹29.3 crore from Chiratae Ventures.
AI for government procurement cuts tender processing time by 60–70%.
Clients: Voltas, Adani, Shapoorji.

TheWedding Company: $2.75 million.
CREST (AI wealth management): $3.1 million from BEENEXT.

The macro picture this week:

$8.15 billion raised in India YTD 2026.
824 equity rounds.
Down 16.5% in capital vs same period 2025.
Down 44% in deal count.

What does a 44% drop in deal count mean?

It means investors are concentrating.

Fewer bets. Larger stakes. Higher bar.

The 2026 Indian startup market is not rewarding stories.
It is rewarding ex*****on backed by numbers.

SolarSquare: 12 years of operating history before this round.
Zepto: doubled revenue before filing.
BazaarNow: 1,800 orders a day before raising growth capital.

Prove it first. Raise second.

The sectors getting capital: climate tech, defence-tech, AI SaaS, Tier 2 commerce.
The sectors struggling: undifferentiated consumer apps without real unit economics.

India now has 127 unicorns and counting.

The next ones won't come from the usual places.

Which sector gets India's next unicorn this year?

Drop your pick below and follow for the weekly report every Monday. 👇

18/06/2026

Disney built a floor that lets you walk through any virtual world.

Not a treadmill.
Not a mat.
A floor made of hundreds of independently moving circular tiles each one adjusting in real time so you stay centred while feeling like you're walking through anything.

A jungle. A spaceship. A battlefield. A Disney park that doesn't exist yet.

It's called HoloTile.

And the most remarkable part isn't even the walking.

It's that multiple people can be on it simultaneously all moving in completely different directions without ever colliding.

Person A walks north. Person B walks east. Person C turns south.
The floor tracks each of them individually.
Adjusts every tile zone independently.
Nobody falls off. Nobody bumps.

The world's first multi-person, omnidirectional, modular, expandable treadmill floor.

Built by Lanny Smoot Disney's most prolific inventor.
The first Disney Imagineer ever inducted into the National Inventors Hall of Fame.
Only the second person from the Walt Disney Company to receive that honour after Walt Disney himself in 2000.

Where does this go beyond Disney parks?

Military and aviation training walk through simulated environments before a real mission.
Medical training surgeons practising procedures while physically moving through a virtual operating room.
Enterprise VR teams sharing a physical-virtual collaborative space.
Architecture walk through a building before a single brick is laid.

And when you pair HoloTile with Google's Android XR glasses announced at I/O in May?

You have the full spatial computing stack.
The glasses complete your eyes.
The floor completes your body.

The spatial internet is not just a screen you look at anymore.
It's a world you physically walk through.

Where would you go first?

Drop it below. 👇

17/06/2026

Apple just handed Google the keys to 2 billion phones.

And most people missed what that actually means.

At WWDC on June 9, Tim Cook in his final keynote as Apple CEO announced that Siri's backend is now powered by Google's Gemini.

Not a partnership. A replacement.

The AI that used to run on Apple's own intelligence layer is now Gemini. On every iPhone. In 40+ languages. Including Hindi, Tamil, and Bengali.

Think about the scale of that for a second.

2 billion active iPhones.
All now running Google's brain.

Gemini is already on 3.6 billion Android devices.
Add 2 billion iPhones and you get 5.6 billion AI-powered devices running one model.

That is the largest AI deployment in history.
Not by training size.
By reach.

What happened to OpenAI in this story?

In 2024, Apple gave ChatGPT the iPhone deal. It was celebrated as a landmark win.

That deal is over.
ChatGPT is now an optional plugin not the default.
Gemini is the default.

The race for AI was never really about who built the smartest model.

It was always about who got on the most devices.

Google just won that race.

For the India angle specifically:
150 million+ iPhones in India.
Gemini now speaks to every one of them in their own language.
For Indian developers: Gemini APIs are now the path to iOS integration.
For Indian brands: Gemini's contextual understanding is the new SEO.

The quiet announcement at the end of Tim Cook's era may be the most consequential tech deal of the decade.

Is this good for users or should we be worried about one AI on 5.6 billion devices?

Drop your take below. 👇

12/06/2026

For the last few years, the AI industry followed one rule:

Send everything to the cloud.

Need an AI assistant?
Cloud.

Need image generation?
Cloud.

Need coding help?
Cloud.

Need a large language model?
Cloud.

NVIDIA's latest move suggests that rule is about to change.

At Computex 2026, NVIDIA unveiled RTX Spark, a new AI-focused computing platform designed to run powerful AI models directly on laptops and desktops instead of relying entirely on remote data centers.

NVIDIA and Microsoft are also working together to make Windows a platform for "personal AI agents" — assistants that can understand context, perform tasks, and operate directly on your device.

The bigger story isn't the hardware.

It's the shift.

For decades, powerful computing moved from PCs to the cloud.

NVIDIA is betting the next generation of AI brings some of that power back.

The future may not be:

Device → Cloud → Answer

It may become:

Device → AI → Action

👇

Would you rather have your AI run in the cloud or directly on your device?

03/06/2026

For years, the AI race has looked like a battle of intelligence.

Who's smarter?

GPT.
Gemini.
Claude.
Grok.

But history suggests something different.

Most people don't know where their electricity comes from.

They just expect the lights to turn on.

AI may follow the same path.

In a few years, many users may never ask:

"Which model am I using?"

They'll simply use whatever is built into:

• their phone
• browser
• operating system
• workplace
• favorite apps

The biggest AI company of the next decade may not be the one with the smartest model.

It may be the one that becomes invisible.

The one that's everywhere.

History often rewards:

Distribution.
Convenience.
Defaults.

Not always the best technology.

The real question is:

👇 What wins in the long run?

Intelligence or Distribution?

01/06/2026