START

The entire vibe coding wave got one thing backwards

Everyone optimized for speed

Prompt in, app out, as fast as possible

But the bottleneck isn't just speed
It's context

Give coding agents better context and everything improves.

Lower cost
Faster execution
Better output

Woz is built for that shift.

A Claude Code plugin that makes coding agents faster, cheaper, and higher performing

Up to 55% cheaper
Up to 40% faster
Higher benchmark scores than Claude Code alone

And the timing matters:

The age of tokenmaxxing is ending.
The age of more measurable ROI is beginning.

🎙️ Ben Collins, CEO & Co-Founder, Woz on Fondo START pod

00:57 Context is the biggest bottleneck
02:11 Why prompt-to-output breaks at enterprise scale
03:28 Apple cracking down on AI-generated apps
05:02 Why local AI models are becoming viable
06:37 The rumored Cursor and xAI acquisition
07:01 Coding agents and the zero-margin problem
09:48 The end of token maxing
12:32 Why AI infrastructure is shifting toward ROI

Check out wozcode.com to learn more.

Your AI agent makes a mistake. Who pays for it?

Not your E&O policy
That assumes a human made the error

Not your cyber policy
That assumes an attacker breached the system

Autonomous AI agents are explicitly excluded from both

Ines Boutemadja discovered this while building AI agents for enterprises

Then procurement started asking for proof the AI was actually covered

Existing policies couldn’t provide it.

So she built Klaimee: the first purpose-build E&O coverage for AI agents

Built for companies deploying AI in healthcare, fintech, and voice & more

Her bet on the future: AI liability is inevitable

‍

🎙️ Ines Boutemadja, CEO & Co-Founder, Klaimee on Fondo START pod 

‍

00:57 "Klaimee is the insurance for your AI agents" 

02:11 Building therapeutic agents and insurance claims agents 

02:48 Existing policies "completely excluded" agentic AI 

03:31 Enterprise procurement blocking deals without coverage 

05:02 Why cyber and E&O fail autonomous AI risk 

06:01 AI liability coverage for agentic system providers 

10:20 First Algerian woman founder in YC 

11:15 "Don't wait for permission" 

14:20 "AI liability is inevitable" 

16:18 Why stealth mode is overrated

Learn more at www.klaimee.ai

‍

AI coding tools are making developers faster than ever

QA still hasn't caught up

Aakash saw it firsthand at Windsurf during the AI coding explosion

Then realized:

QA is probably a bigger bottleneck than coding itself

So he built Canary

An AI QA engineer that reads your source code, maps changes to real user flows, spins up remote browsers, and drops video recordings directly into the PR

15 minutes instead of days

Now the workflows are getting even crazier

One customer already runs a fully autonomous loop:

Linear ticket → Cursor background agent → PR → Canary QA → Merge

The engineer just reviews the outputs before approving the change

Code generation is becoming autonomous

QA has to become autonomous too

🎙️ Aakash Mahalingam, Co-founder & CEO, Canary on Fondo START pod 

‍

00:15 What Canary does 

0:57 Spinning up remote browsers to test 

1:42 User flow validation vs file reviews 

2:29 How this changes developer workflows 

3:12 Lessons from Windsurf and Cognition 

3:43 QA is a bigger bottleneck than coding itself

4:08 Pivoting inside YC from regression to feature testing 

5:02 Applying to YC five times 

6:11 Background agents opening PRs autonomously 

6:37 "They didn't write a single line of code to merge that"

‍

Check out runcanary.ai to learn more. 

The “buy a list and blast emails” era is ending

Not because outbound stopped working

Because the data got smarter than the emails

Kathryn Wu and her co-founder built Openmart as an SMB intelligence layer:
Verified owners
Decision-maker contacts
Google reviews
Website quality
Location intelligence & more

Now they've launched "OpenClaw for Sales"

A conversational outbound workspace built on top of that data

No spreadsheet cleanup
No disconnected enrichment tools
No giant table views

Just ask the database what you need

DoorDash uses Openmart for regional planning
Whatnot uses it to identify high-quality sellers
Clay customers use the data directly inside outbound workflows

Try it out at openmart.com

🎙️ Kathryn Wu, Co-Founder of Openmart on Fondo START pod

‍

01:13 Openmart’s SMB intelligence database and Openclaw for Sales

02:16 The shift from giant lead lists to AI qualification and scoring

03:43 Why SMB data differs from LinkedIn-centric sales platforms

05:57 How modern GTM teams use multi-channel outbound workflows

06:39 The highest-value qualification signals in SMB sales

08:25 Openclaw for Sales and conversational outbound workflows

09:02 How DoorDash uses Openmart for regional planning

11:28 Deduplication and the hidden pain of outbound infrastructure

12:16 “The future is qualification and scoring”

14:08 Why proprietary data is the moat behind AI outbound

Learn more at openmart.com

A drug was approved in China

Then it failed clinical trials in the US

No one could figure out why

The answer was hiding in the gut microbiome
Anto Bio’s model helped identify it

99% of the genes in your body aren't yours 
They belong to your microbiome
Over two-thirds of drugs are heavily affected by how your gut processes them

But the microbiome data was too noisy. Too messy. Petabytes of it just sitting there

Anto Bio's mission: make the gut microbiome computable for the first time

They're building foundation models that help pharma design drugs that work for more people

🎙️ Arvid Gollwitzer, Co-Founder, Anto Bio on Fondo START pod

1:18 What Anto Bio is building and why it's novel

2:10 Why this didn't exist before

3:05 99% of your genes aren't yours

4:00 Why pharma ignoring the microbiome is a huge problem

5:04 The microbiome is modifiable

6:30 Modeling microbiome differences across populations

7:00 Drug approved in China, failed in the US, and how Anto found out why

8:02 How many drugs are affected by microbiome metabolism

10:00 GLP-1s and the gut microbiome connection

12:42 The biggest unsolved variable in drug development

Learn more at www.anto.bio

Most people think they need more motivation to work out

They need less friction

But training plans assume your life stays static

You travel
It rains 
You have 20 mins between meetings

People don't quit because they lack discipline

They quit because adapting the plan is harder than the workout itself

HYBRD brain can help

An adaptive AI agent that adjusts your training to whatever life throws at you

Built by ultramarathoners, Ironman finishers, and a founding engineer who won a 100-mile cycling race

🎙️ Matthew Ruiters, CTO & Co-Founder, HYBRD on Fondo START pod

02:01 Connecting every wearable into one training system

02:42 Using sports science to measure total training load

03:07 How former Whoop teammates started building HYBRD

03:50 Why “serious athlete” is really a mindset

04:49 The idea behind HYBRD Brain

05:18 Adapting workouts around travel, weather, and hotel gyms

06:13 The consistency insight from 895,000 athletes

10:44 Why athletes wanted adaptation, not just tracking

11:35 Why friction stops founders from staying healthy

14:03 How AI tools are changing software engineering

learn more at www.hybrd.com

Every voice AI has the same problem

The moment you stop talking, it starts talking

But humans don’t work that way

In real conversations, pauses don’t always mean interruption. Sometimes they mean: keep going

Current AI misses this completely.

Amit Yadav spent years building real-time AI systems at Meta Reality Labs for Ray-Ban smart glasses

Now he’s building Fern: a sales copilot that assists teams live during meetings.

Not after the call.
During it.

The hard problem isn’t retrieval anymore.

It’s judgment.

Because the future of AI won’t be defined by smarter answers alone.

It’ll be defined by systems that understand how humans actually communicate.

🎙️ Amit Yadav, CEO & Co-Founder of Fern, on Fondo START pod

‍

01:24 “Imagine your sales rep hears a prospect question and answers appear instantly.”

01:55 The hardest AI problem: knowing the right time to help

02:09 “Our in-house model beats human judgment, ChatGPT, and Gemini.”

02:46 Building realtime AI systems for Ray-Ban Meta smart glasses

03:57 Fern’s desktop copilot listens and assists during meetings

04:34 The industry shift from prompted AI → proactive AI assistance

05:28 Replacing Slack searches, docs, and tabs during live calls

07:10 “The most exciting thing is AI helping you without asking for it.”

11:09 Why silence in human conversation doesn’t always mean interruption

14:15 The need for a realtime social-awareness layer between humans and LLMs

‍
Check out getfern.ai

Speed to lead sounds like a sales tactic

It's actually the whole game

Ruben Harris knows this better than anyone

He built a 300-person call center across 3countries to reach 1M people a month 

...and saw exactly where the ceiling was

The problem isn't effort. It's physics. 

A working-class student fills out a form at 10 PM. Nobody's staffed at 10 PM

So 90% of leads go cold
Not because they weren't interested
Because nobody got back to them fast enough

Outrival's AI agents respond in under a minute

They call, text, email, and follow up

One client made $10M in new revenue from leads they already had

The best outbound AI isn't replacing humans
It's doing the work humans never had time to do

🎙️ Ruben Harris, Co-founder & CEO, Outrival on Fondo START pod
‍

01:12 Why Outrival focuses on outbound AI
03:18 Building a 300-person call center
05:27 Why speed-to-lead changes everything
08:52 Aged leads as hidden revenue
09:57 100,000 calls in an hour
14:11 Turning one school’s leads into $10M revenue
18:43 Enterprise sales advice for AI founders
24:48 Pricing digital workers like employees
27:56 From cello to Wall Street to Silicon Valley
50:14 Building startups in the AI era

Check out www.outrival.com

"Turn every interaction into a long-term relationship"

The way to make software faster was always: make the code happen faster

AI solved that 

Cursor, Claude Code, Replit...

Building is nearly instant now
So what's slow?

Knowing if it actually works.

When AI writes your feature, someone still has to answer: did it break everything else?

That's the new bottleneck. Not creation. Validation
Jack Brown is building Lark to solve exactly this. 

An AI QA engineer that writes tests, keeps them updated, and continuously validates

There's the building, and then there's the thing that lives on after the building

Lark is focused on that second half
‍
🎙️ Jack Brown, Cofounder, Lark on Fondo START pod

00:15 Building an AI QA engineer
01:56 AI dramatically increased development velocity
02:15 The bottleneck becomes verification
02:25 Cursor writes the code. Lark writes the tests.
03:24 Acceptance criteria → automated validation
04:15 Non-technical founders building software with AI
05:35  Lark agents autonomously defining tests
06:31 Everything is just validation
07:56 Fully agentic software creation
08:53 Why testing becomes more critical in AI-generated codebases

Visit getlark.ai to learn more

Google used to surface the PDF of a law if you searched its exact name.

Now it surfaces consultancies advising on it.

And LLMs have gotten worse at this too. As they’ve tried to cut costs and focus on the information 99.99% of users want, long-tail edge-case data has become much harder to find.

Natalie Aresta-Katz was at a big law firm. One of their biggest clients wanted to track new and proposed laws in 45 countries that didn’t publish in English.

They hired lawyers in every single country. Tried probably 15 tools on the market. Ran manual searches.

Nothing worked well.

Now she's building the tool she wished she had.

One workflow went from up to 13 paralegals and around 120 hours a month to one user and about an hour and a half.

‍

🎙️ Natalie Aresta-Katz, CEO & Founder of Regbase, on Fondo START pod (full ep in comments)

‍

00:20 “Regbase was really born out of a place of necessity.”

01:32 Tracking new and proposed laws across 45 countries

02:00 Why existing tools failed

02:43 The “Wizard of Oz” backroom behind regulatory databases

03:26 Why long-tail legal data is getting harder to find

04:23 Dropping into YC after Andrew said, “an MBA isn’t really school”

05:37 Finding unspent state-level funding for school safety upgrades

07:05 From 120 hours to about 1.5 hours

07:38 How Regbase uses LLMs, specialty search, and link-following

08:33 Why legal work may shift from hourly billing to output

09:38 Searching laws in languages teams don’t speak

10:27 Why Regbase unlocks more opportunity for law firms

11:00 Who should book a demo: government vendors, corporate lawyers, AI law trackers, employment lawyers, and teams tracking high-volume regulatory change

visit regbase.com to learn more