We're at the halfway point of 2025, and AI is evolving faster than ever.
In this solo episode, I share Untapped Ventures’ first Request for Startups. It’s a detailed breakdown of the core areas where we’re actively looking to invest. If you're a founder building in AI or exploring ideas, this is the roadmap for what we want to fund next.
I walk through the key themes that reflect our thesis across both Agentic AI and Convergence AI. These are the ideas we believe will drive the next generation of breakout companies.
Inside the episode:
• Why service-as-software will replace traditional SaaS
• How full-stack agentic platforms are reshaping vertical industries
• The rise of outcome-based pricing and what it unlocks
• Where lean AI teams are gaining massive traction
• Why we’re excited about AI converging with biotech, neurotech, energy, defense, and more
• The profiles of founders we love to back
At Untapped, we invest from stealth and pre-salt through seed, writing checks between $500K and $2 million. Whether you're actively raising or still in the idea maze, we want to meet you.
Let’s connect!
George Bandarian: https://www.linkedin.com/in/georgebandarian
Explore further:
1. If in Stealth, let’s chat: https://bit.ly/UntappedE18podS
2. If raising, pitch us: https://bit.ly/UntappedE18pod
Is your AI startup built to survive the next wave?
In this episode of Agentic, Will Manidis, the co-founder of ScienceIO, shares how he built for resilience from day one developing LLMs for healthcare and the frameworks that helped him scale before the world caught up.
He shares how local model training on office gaming PCs laid the unlikely foundation for ScienceIO’s reach across millions in digital health.
We also discuss:
• why OpenAI’s API continues to dominate Gemini
• the hidden fragility behind $500M fast scaling AI companies
• Will’s belief that founders must prove value before overbuilding for scale
Not often do you get to learn from someone who built an LLM before ChatGPT. Tune in now.
If you’re in the early innings of building something in AI, especially if you’re figuring things out from first principles—click here, we'd love to hear from you.
Connect with us here:
1. Will Manidis- https://www.linkedin.com/in/willmanidis/
2. George Bandarian- https://www.linkedin.com/in/georgebandarian/
3. Untapped Ventures- https://www.linkedin.com/company/untappedventures/
00:00 Trailer
01:41 Introduction
02:45 Thiel Fellow
07:40 Pre-ChatGPT LLMs
10:27 Biggest lessons
12:37 Contrarian takes
13:56 Difficult to sell
15:59 First exit
16:38 Getting world-class talents
23:20 Risk mitigation strategy
25:48 Defensibility and moats
37:10 Basic MVP
44:01 Q&A
50:35 Outro
Most no-code tools give you an MVP. Woz wants to give you a business.
On this episode of Agentic, I was joined by Ben Collins and Brad Eckert, co-founders of Woz, an AI platform that lets anyone build real, scalable web and mobile products with zero coding. Think Shopify, but for launching operational SaaS startups.
They break down:
• Why shipping your MVP is just 1% of the journey—and what the other 99% really looks like
• The difference between vibe coding, AI engineering, and real startup durability
• What they learned from going through Y Combinator (twice), taking 80+ investor meetings, and raising a breakout seed round.
• Why automation means more software businesses and not fewer jobs
00:00 Trailer
01:07 Introduction
02:29 Empowering entrepreneurs
04:27 Woz's Vision
10:32 Success stories
13:38 The origin story
15:27 AI code generation evolution
17:29 Technical insights
19:44 The Role of AI
30:31 Job markets
31:20 Entrepreneurial mindset shift
32:56 Business building challenges
34:54 Creating a successful platform
37:45 Y Combinator experience
43:41 Fundraising insights
49:29 Advice for AI Entrepreneurs
56:31 Closing
What happens when the former Chief Scientist at Salesforce builds a search engine for enterprises?
In this episode, Richard Socher (CEO) and Bryan McCann (CTO) of You.com reveal how they’re reinventing enterprise search and agentic workflows with their latest product, Ari—a deep research assistant that goes far beyond fast facts.
We explore:
00:00 Trailer
01:56 Introduction
03:41 Richard and Bryan's Meeting
09:52 Foundations of search
12:33 Deep research agent
13:49 How You.com stands out
16:03 Custom agents
18:15 ARI use cases
19:43 Going slow to go fast
24:00 Models under the hood
26:00 AGI workflows
29:11 Adoption
33:34 Complicated workflows
36:45 AGI - are we there?
46:20 How to compete as a startup
53:32 Being right
56:23 Outro
Everyone has opinions about founder equity, funding timelines, and valuations—but Peter Walker holds the data.
In this special live Agentic session, I sat down with Peter, Head of Insights at Carta, to unpack what’s actually happening beneath the surface in early-stage AI startups.
Peter’s team has a privileged vantage point—cap tables, valuations, and hiring data across 45,000 startups and 83% of U.S. unicorns.
We explore:
If you’re raising, hiring, or advising an AI startup in this market, this one’s for you.
Connect with us here:
00:00 Trailer
01:26 Introduction
02:44 Untapped Ventures
08:39 Carta
10:34 Early Equity
25:05 Venture Context
34:47 AI Story
51:04 Questions
54:48 Outro
I brought Ben Collins of Woz and Dakotah Rice and Tushar Nair of Harper for a special episode of Agentic–our AGM!
Ben’s Woz lets non-technical founders build scalable apps with no engineering team—think Shopify for software while Dakotah and Tushar’s Harper quotes 476 complex insurance deals with voice AI, turning family brokerage chaos into a data goldmine.
On the docket:
Connect with us:
As AI races forward, one question looms large—who decides what’s right and wrong?
Behind the breakthroughs and multi billion-dollar valuations, a power struggle is unfolding. Big Tech is quietly shaping AI policy in the shadows, and governments are stuck playing catch-up.
In this episode, I speak with Wendell Wallach, a foremost expert on AI ethics and governance, to break down the challenges and power dynamics steering the course of AI.
Co-author of Moral Machines and a key figure at Yale and Carnegie Council, Wendell has spent decades thinking about how to align AI with human values.
We cover:
• why rogue superintelligence isn’t the biggest threat
• how Big Tech is quietly shaping AI policies
• AGI vs. ASI: When AI starts making its own rules
Connect with us here:
1. Wendell Wallach: https://www.linkedin.com/in/wendell-wallach-246ba69/
2. George Bandarian: https://www.linkedin.com/in/georgebandarian/
3. Untapped Ventures: https://www.linkedin.com/company/untappedventures
00:00 Trailer
00:55 Introduction
01:59 AI uplifting humanity
04:09 Moral machines
09:58 Big tech oligopoly
16:00 Cold War
18:59 AI will take over the world?
29:57 Nefarious purposes
34:32 Trade-offs and safeguards
38:11 Data security
41:00 Finding product market fit
43:25 Human dangers
47:46 Outro
He built AI-powered lunar rovers at NASA.
He was in the room when GPT launched.
Now, Logan Kilpatrick is helping developers push the boundaries of AI at Google DeepMind.
In this episode, Logan shares his front-row view of AI’s most pivotal moments—and how they hint at the next paradigm shifts.
We also get into:
• inside Open AI on Launch day
• how Google DeepMind’s Gemini 2.0 Flash is flipping the script on AI economics
• the open-source vs. proprietary AI debate, and
• why AI’s next frontier is deeply integrated, multimodal agents
Plus, Logan unpacks his insights as an investor—where he sees opportunity in AI, what founders keep getting wrong, and why the most successful startups won’t look like today’s incumbents.
00:00 Trailer
01:19 Introduction
02:41 Inspired by Floppybird
03:53 OpenAI Developer Relations
08:55 Where AI is heading
14:41 Google DeepMind
17:47 As an AI founder, B2B and GSuite customer
21:29 Better model, low cost
27:12 Agentic AI and Google Gemini
29:32 AGI and DeekSeek
35:55 Investor and AI founder perspectives
44:43 What's next from Google
48:38 Outro
Bigger AI models dominate the headlines—but what if the real breakthrough lies in making AI smaller?
In this episode, I sit down with Ramin Hasani, Alexander Amini, and Daniela Rus—who are at the helm of Liquid AI—to discuss how their approach challenges conventional architectures and unlocks new frontiers for AI deployment.
Also on the docket:
• Ramin explains how insights from a microscopic worm led to a novel AI model.
• Alexander breaks down why Liquid AI operates efficiently on local devices—without compromising modality.
• Daniela exposes a critical flaw in today’s AI incumbents and why efficiency is the next major battleground.
We also unpack all things DeepSeek—its implications for OpenAI, Meta, and enterprises scaling private AI.
00:00 Introduction
02:03 Meet the Co-Founders
03:01 The Birth of Liquid Neural Networks
06:28 Applications and Impact of Liquid AI
09:38 The Worm Model and Its Significance
16:05 Mathematical Foundations and Breakthroughs
24:29 Scaling AI for Real-World Applications
28:05 Edge Computing and AI
29:58 Future Prospects and Use Cases
31:36 The Commoditization of AI Models
32:22 Liquid AI: Reducing Intelligence Costs to Zero
32:53 Diverse Applications of AI
33:43 Case Studies and Real-World Examples
38:05 The Impact of Deep Seek on the AI Ecosystem
44:13 Advice for Aspiring AI Founders
45:40 The Importance of Technical and Non-Technical Collaboration
51:03 Ramin's Journey from Scientist to Entrepreneur
53:31 Liquid AI's Vision for 2025
55:10 Conclusion and Future Prospects
For years, $7.2B in U.S. government-backed neuroperformance research sat untouched—proven cognitive enhancements with no commercial application.
David Bach saw the gap. Optios is bridging classified IP with market demand, transforming military-grade neuroscience into real-world performance gains.
The upside? Monumental. The imbalance? Even bigger.
We dive into:
• how declassified DOD research is unlocking a new frontier in neurotech
• the hidden psychology behind winning investor buy-in
• what $600M+ in exits taught him about product-market fit
Connect with us here:
1. David Bach:https://www.linkedin.com/in/david-bach-md-0b81b29/
2. George Bandarian:https://www.linkedin.com/in/georgebandarian/
3. Untapped Ventures:https://www.linkedin.com/company/untappedventures
00:00 Trailer
01:20 Introduction
01:46 David Bach
04:46 Journey to Harvard
10:26 Founding Optios
16:08 Tech Transfer and Market Research
22:09 The Future of Neuro-Performance
28:33 Key Lessons for Founders
33:34 Fundraising Insights
37:29 Closing Thoughts
AI agents will drive the next wave of automation—but only if they stop breaking.
In this episode, I’m joined by Alex Reibman, Adam Silverman, and Braelyn Boynton, the powerhouse team behind Agency.
Together, they’re tackling the toughest challenges in AI agent development, from failure recovery to debugging and slashing operational costs.
In this episode:
• Alex reveals why today’s AI agents break in unpredictable ways—and how Agency is solving it.
• Adam unpacks OpenAI’s latest agent release and its ripple effects across the ecosystem.
• Braelyn explains how they’re standardizing agent development, lowering barriers for builders everywhere.
Let’s dive in.
Connect with us here:
1. Alex Reibman- https://www.linkedin.com/in/alex-reibman-67951589/
2. Adam Silverman- https://www.linkedin.com/in/adamsil/
3. Braelyn Boynton- https://www.linkedin.com/in/braelyn-boynton-75911974/
4. George Bandarian: https://www.linkedin.com/in/georgebandarian/
5. Untapped Ventures: https://www.linkedin.com/company/untappedventures/
00:00 Trailer
01:31 Welcome to the Agentic Podcast
02:00 Meet the Founders of Agency AI
03:03 The Origin Story of Agency AI
03:31 Challenges and Solutions in AI Agent Development
04:24 Understanding the AI Ecosystem
04:53 Backgrounds of the Founding Team
09:55 The Evolution of AI Agents
18:46 Enterprise Use Cases for AI Agents
21:22 Innovative Applications of AI Agents
29:25 Building an Agentic Process for VC
32:52 Task Generation and YAML Prompting
33:23 Modifying and Scaling Agents
35:34 AgentOps and Prompt Engineering
39:23 Fundraising Journey and Lessons
43:53 Future Vision and Developer Ecosystem
46:47 Advice for Founders and Engineers
54:28 Hackathon Strategy and Community Building
01:02:36 Concluding Thoughts and Final Advice
AI is a platform shift unlike any other, with far-reaching implications for every startup.
In my recent webinar with the Founder’s Bay community in Silicon Valley, hosted by Mariane Bekker, I explored why AI should be the core focus for founders aiming to build scalable, high-impact companies.
Key takeaways:
• there’s hardly an application that AI can’t improve
• AI startups are exiting faster—7 years on average vs. 13 for others
• building in AI is your best shot at hitting $100M ARR
• it’s not a zero-sum market—horizontal and vertical plays can both win
• focus on solving high-value problems using AI to attract funding and top-tier talent
In this episode, I share my views on how you could build a leading AI startup that attracts funding, customers, and talent in 2025.
Connect with us here:
1. George Bandarian: https://www.linkedin.com/in/georgebandarian/
2. Untapped Ventures: https://www.linkedin.com/company/untappedventures/
00:00 Trailer
03:19 Introduction
05:53 State of AI
11:55 Ideation
15:48 Agentic AI
26:32 Validation
29:06 Traction
31:48 Recruiting
33:00 Fundraising
35:28 Value-add
37:15 Takeaways
39:01 Q&A
55:28 Outro
The insurance industry has a 400,000 agent shortage.
I sat down with Harper's founders, who use agentic AI to deliver enterprise-grade service at 1/10th the cost.
Dakotah Rice and Tushar Nair revealed how they're leveraging the $85B E&S market:
• voice AI running quotes to claims
• why selling AI tools failed, but becoming a broker worked
• how they're hitting software margins in a services business
As insurance carriers exit Florida and California, Harper couldn’t have come at a better time.
Connect with us:
Dakotah Rice: https://www.linkedin.com/in/dakotahrice/
Tushar Nair: https://www.linkedin.com/in/tushar-nair-b08988222/
Harper: https://www.linkedin.com/company/harperinsure/
George Bandarian: https://www.linkedin.com/in/georgebandarian/
Untapped Ventures: https://www.linkedin.com/company/untappedventures/
00:00 Trailer + Introduction
04:33 Dakota's Journey to Founding Harper
05:18 Tushar's Background and Expertise
06:06 The Genesis of Harper
06:38 Challenges and Opportunities in AI-Driven Insurance
13:08 The Evolution of Harper's AI Systems
13:21 Technical Insights into Building AI Agents
16:01 The Future of AI in Insurance
17:36 Advice for Aspiring AI Founders
19:55 Focusing on the ENS Market
22:05 Scaling and Expanding Harper's Services
25:37 Innovative Sales Solutions for Insurance
26:14 Customer Acquisition and Marketing Strategies
27:42 Voice AI: Revolutionizing Customer Interaction
30:15 Leveraging AI for Enhanced Customer Service
33:02 Future of AI in Business Operations
35:49 Fundraising and Business Growth Insights
40:39 Advice for Founders: Building and Scaling Startups
46:16 Co-Founder Dynamics and Relationship
50:08 Conclusion and Final Thoughts
Only 8 companies have shipped AI Agents from production. In this episode of Agentic, I brought together 3 of them, to reveal how they crossed this chasm:
1. Aria Attar (TensorStax) shows how they're tackling the 1:250 data scientist shortage with agents that drop into existing cloud environments.
2. Joe Stolte (Daily.ai), fresh off multiple exits including a $62.5 M acquisition, reveals how they turned 1B newsletter interactions into $1.2M in ARR in just 14 months.
3. Jacky Wong (Nullify) explains how they're helping enterprises where security teams are outnumbered 4:1 by developers.
Key insight? While the AI market races toward $733.7B by 2027—
Solving for deployment and LLM reliability > focussing on model capabilities.
Let’s dive right in.
Connect with us at:
1. Jacky Wong: https://www.linkedin.com/in/wongja/?originalSubdomain=au
2. Aria Attar: https://www.linkedin.com/in/ariaattar/
3. Joe Stolte: https://www.linkedin.com/in/josephstolte
5. George Bandarian: https://www.linkedin.com/in/georgebandarian/
6. Untapped Ventures: https://www.linkedin.com/company/untappedventures/
00:00 Trailer + Introduction
02:26 Agentic AI Overview
06:49 Meet the Founders
10:42 Deep Dive into AI Security
20:01 Building Agentic Models
25:59 Deep Reasoning and Inference in AI Models
26:19 Building and Training Baby Agents
27:37 Trust and Acceptance in AI Solutions
28:49 Lessons Learned and Advice to Founders
29:40 Common Mistakes in LLM Development
33:32 Customer Acquisition Strategies
36:57 Fundraising Insights and Challenges
42:38 Driving Acceptance and Trust in AI
44:37 Reconciling AI as a Deflationary Technology
46:56 Removing Human Evaluators in Development
With $90B poured into AI over the past year, where is the focus going to be in 2025? I spoke to four top investors to find out.
1. Yipeng Zhao (Embark Ventures) believes AI offers a one-year head start. Fast GTM wins. Slow movers risk irrelevance.
2. Hari Arul (Khosla Ventures) has stopped backing co-pilots altogether. His take? Human-in-loop models won’t survive past 2026.
3. James Shecter (Fika Ventures) says enterprises need the illusion of control over their AI workflows.
4. Nicole Toussaint (TenOneTen Ventures) sees the “Scopely moment” for AI agents approaching fast — and soon, market entry will outweigh engineering breakthroughs.
Tune in now.
Connect with us at:
1. Yipeng Zhao: https://www.linkedin.com/in/yipengzhao/
2. Hari Arul: https://www.linkedin.com/in/hariarul/
3. James Shecter: https://www.linkedin.com/in/jamesshecter/
4. Nicole Toussaint: https://www.linkedin.com/in/nicole-toussaint-2874a1106/
5. George Bandarian: https://www.linkedin.com/in/georgebandarian/
6. Untapped Ventures: https://www.linkedin.com/company/untappedventures/
While 92% of Fortune 500s realize the surplus alpha that can be created from leveraging AI, only 25% of SMBs realize the same.
I sat with 6 founders who are building Agentic AI solutions: Jacky Wong (Nullify), Aria Attar (TensorStax), Max Ahrens (Maihem), Saurabh Sharma (You.com), Jonas Diezun (Beam AI), and Ashar Rizqi (Bounti.ai).
The key highlights from our conversation were:
- OpenAI’s function calling breaks down when scaling requires consistent outputs and error handling—just ask Jacky or anyone building enterprise agents.
- MongoDB could face disruption as AI agents take over infrastructure orchestration—Aria flagged it, and I agree.
- Calling agents "AI workers" simplifies adoption for hesitant enterprises—a shift that doubled Jonas Diezun’s sales.
- Agentic AI scales across verticals in months, not years—Ashar says the old SaaS scaling playbook is obsolete.
- Max at Maihem points out that building functional agents hinges on managing the edge cases that crash your system.
Let’s dive right in.
Connect with us here:
1. Jacky Wong- https://www.linkedin.com/in/wongja/?originalSubdomain=au
2. Aria Attar - https://www.linkedin.com/in/ariaattar/
3. Max Ahrens - https://www.linkedin.com/in/maximilianahrens/?originalSubdomain=uk
4. Saurabh Sharma - https://www.linkedin.com/in/saurabhs/
5. Jonas Diezun - https://www.linkedin.com/in/jonasdiezun/
6. Ashar Rizqi - https://www.linkedin.com/in/ashar-rizqi-6614487/
7. George Bandarian - https://www.linkedin.com/in/georgebandarian/
8. Untapped Ventures - https://www.linkedin.com/company/untappedventures/
00:00 Trailer
01:26 Welcome to the Agentic Podcast
02:21 Founder Panel Introduction
02:41 Founders' Backgrounds and Insights
10:08 Defining Agentic AI
20:00 Building Effective AI Agents
26:08 Challenges in Early AI Development
26:21 Human in the Loop: Enhancing AI Adoption
26:58 Debating the Importance of Evaluations
27:39 Technical Insights and Mistakes
29:07 Impact of Agentic Paradigm Shift on Infrastructure
29:52 Customer Adoption and Market Strategies
31:21 Go-to-Market Strategies and Early Stage Focus
39:18 Horizontal vs Vertical Approaches in AI
47:39 Final Thoughts
#ai #tech #podcast #vc
As we transition into agentic models, it’s evident there’s a blend of optimism and caution.
In this panel discussion, I was joined by Jon Chu (Khosla Ventures), Pranav Reddy (Conviction), Andrew Brackin (Gradient), Arash Afrakhteh (Pear VC), Alejandra Vergara (Bee Partners), and Nahim Nasser (Georgian) as we explored:
• why agentic AI will drive systemic shifts across various industries
• the need to define scalable, high-impact use cases for agent-driven solutions
• shifting dynamics of human-machine interaction
• how agentic AI is advancing in parallel with LLMs
• why regional disparities and emerging foundation models could profoundly shape future progress, and
• how value will accumulate for vertical companies beyond the API
Let’s jump right in.
00:00:00 Trailer
00:02:13 Introduction
00:11:47 Current State of AI
00:21:00 Investment Insights and Startup Pitfalls
00:30:49 Vertical vs Horizontal
00:42:09 Evaluating Business and Team Dynamics
00:51:19 Different ways to look at AI
00:58:06 The AGI Landscape
01:01:17 Outro
Connect with us here:
1. Jon Chu- https://www.linkedin.com/in/jchu/
2. Pranav Reddy- https://www.linkedin.com/in/pranavgreddy/
3. Andrew Brackin- https://www.linkedin.com/in/andrewbrackin/
4. Arash Afrakhteh- https://www.linkedin.com/in/arasha/
5. Alejandra Vergara- https://www.linkedin.com/in/alejandra-vergaral/
6. Nahim Nasser- https://www.linkedin.com/in/nahimnasser/
To me, agentic is intent meeting capability—systems that act autonomously, adapt intelligently, and drive purposeful outcomes.
Our guests on this episode—Jonathan Corbin, Sami Shalabi, and Eugene Mann—embody that. They're pushing AI agents forward for real-world enterprise use cases with Maven AGI.
We discuss:
• solving complex challenges with AI agents
• empowering enterprises with cutting-edge tools, and
• why agentic AI will create new roles for humankind
Let’s get into it.
Connect with us here:
1. Jonathan Corbin- https://www.linkedin.com/in/jcorbin01/
2. Sami Shalabi- https://www.linkedin.com/in/sshalabi/
3. Eugene Mann- https://www.linkedin.com/in/eugenemann/
4. George Bandarian- https://www.linkedin.com/in/georgebandarian/
5. Untapped Ventures- https://www.linkedin.com/company/untappedventures/
