Podcast Summary: This Week in Startups — "Compliance Startup Scandal... Is Delve Guilty?" | E2266
Date: March 24, 2026
Host: Jason Calacanis (JCal)
Co-host: Alex Wilhelm
Guests:
- Elizabeth Yin, Hustle Fund
- Ryan Madhavi, Seal
- Seb Sheng, Brick
- Gavin Zainz & Pranav Ramesh, LeadPoet (Bittensor subnet 71)
Episode Overview
This episode delves deep into the alleged compliance fraud at Delve, a Y Combinator (YC)-backed startup in the automated compliance space. The hosts also explore the broader impacts of AI on startups, the evolving responsibilities of venture capitalists, governance in high-growth companies, and new frontiers in decentralized and agentic tech ecosystems (with discussions about TAO, Bittensor, and Web3 applications). The show is packed with insights for founders, investors, and anyone following tech’s wildest frontiers.
Key Discussion Points & Insights
1. The State of Early-Stage Startups in the AI Era
- AI is changing everything:
- Elizabeth Yin reveals how internal operations at Hustle Fund now leverage AI tools for coding, automation, and investing workflows.
- Quote: “Everybody can be spinning up these apps over a weekend...the amount of capital you need for software has gone down dramatically...companies get faster growth and so they need to pay for their compute bill. So we're still needed on that level.” — Elizabeth Yin [04:33]
- Barriers are lower, but competition is fiercer:
- Founders can now upskill quickly, enhancing abilities in previously weak areas.
- This democratization raises the bar for everyone; those not adapting are being left behind.
- Quote: “If you can be good at the dozens of things it takes to do the chores at a startup...that's a big breakout.” — Jason Calacanis [09:09]
2. Startup Governance, Diligence, and the Culture of “Hacking” Reality
-
Culture of ‘breaking rules’:
- Silicon Valley’s hacker mentality sometimes bleeds into dangerous territory.
- Investors like YC even ask applicants to tell stories about when they’ve hacked something, blurring the line between creative problem solving and fraud.
- Quote: "There's a difference between hustling and faking it...but 500 boilerplate reports, nearly half of which are basically identical...if proven true, this is gaming the system." — Ryan Madhavi [17:09]
-
Due diligence realities:
- At early-stage, investors often can’t get into “the weeds" and must trust founder reporting, which makes the ecosystem fragile to bad actors.
- Quote: “It’s a trust-based environment... You rely on the trust of the founder to send you a monthly, quarterly update... but we do diligence.” — Jason Calacanis [22:18]
-
Red/pink flags:
- Founders exaggerating customer/client lists in decks—a major reason for “dinging” or rejecting startups.
- The line between exaggeration and lying is fuzzy, but clarity and transparency are essential:
- Quote: “You need to be crystal clear about the state of everything. If somebody is not actually your customer...we will ding for that because that suggests future behavior in exaggerating in other ways.” — Elizabeth Yin [26:45]
3. The Delve Scandal: What Happened?
[13:20-31:05] Main Discussion
-
Background: Delve, an AI compliance startup, is accused of fabricating up to 500 boilerplate SOC2 reports, swapping logos, and passing them off as completed work — a statistically improbable scenario suggesting fraud.
-
The Leak:
- An anonymous Substack (“deep delver”) analyzed leaked data and identified evidence of copy-pasted, non-personalized compliance reports and fake auditor networks.
- Major Delve customers potentially impacted include high-profile brands and even a public company.
-
Implications:
- Criminal risk: If proven, actions could lead to criminal charges.
- Second-order effects: Delve’s clients could be put at risk with their own customers due to faulty compliance documentation.
- Industry trust shaken: The trust-based compliance automation sector faces industry-wide scrutiny.
-
Quotable:
- "If this is true...and they just changed the logo on [SOC2 reports] and it's like AI slop or it's cut and pasted...that would be, you know, like definitionally fraud." — Jason Calacanis [17:56]
- “Whoever put this investigation piece together, bravo. Level of forensic detail is impeccable.” — Ryan Madhavi [14:55]
Why Did This Happen?
- Overreliance on trust and “diligence by proxy.”
- Hype cycles and urgency (“you’re either in or out”) led late-stage investors to forgo deep diligence, assuming previous investors had done so.
4. How Can Investors/Customers Protect Themselves?
-
Investor side:
- Early investors often trust what founders report, but as rounds/prices go up, diligence should deepen.
- Relying on another investor’s diligence (without reviewing notes or conducting one’s own) is a “cardinal sin.”
- Governance matters:
- Having boards and board meetings, particularly at Series A and beyond, is essential for oversight and healthy spend/go-to-market plans.
- Startups resisting investor involvement (e.g., YC’s “no board seats” advice) may create risk blindspots.
-
Customer side:
- Ask for credentials: Customers should request actual auditor certifications/proofs.
- Demand transparency: Request product demos—“a 10–30 minute demo would have exposed a lot of so-called features that don’t even exist” [27:54].
5. The Modern Go-to-Market Arms Race: Billboards, Donuts...and Brandwashing
- Delve's sales hacks:
- Lavish non-product “growth hacks” (donuts, doormats, billboards) used to attract clients or mollify skeptics.
- Reciprocity as a tactic is fine, but over-indexing on such “hacks” while shipping a weak product is a bad sign.
- Quote:
- “Dell went around SF handing out donuts...Donuts on the front end and donuts on the back end [to smooth things over].” — Alex Wilhelm [34:44]
- Brand-washing:
- Using logos of well-known customers is common, but if exaggerated or faked, it's a huge red flag (see: Catch Me If You Can analogy).
- True “social proof” is good, but must be legitimate.
6. New Frontiers: Energy AI and Web3/Agentic Ecosystems
Interview: Seb Sheng, Brick [45:28–56:05]
- Brick’s Pitch:
- AI hardware/software blend for energy optimization in buildings and data centers.
- Results: 15%-35% energy savings; $30-50K annual contract value typical.
- Challenge:
- Biggest barrier is scaling pilots to wide, full deployment—slow-moving incumbents and cost of deployment.
[58:51–end] Jason’s Deep Dive: TAO, OpenClaw, and Bittensor
-
The Web3/AI Mash-up:
- Bittensor’s network—subnets are like decentralized “app stores” for AI/data services, where contributors earn tokens (TAO) by running agents or providing data.
- LeadPoet: Subnet #71, is a decentralized lead-gen service relying on miners and validators to provide verified business leads at rock-bottom prices (now just 3-5 cents/lead).
- Disruptive angle:
- Minimizes costs by open global competition, incentivizing high-quality, low-cost data and breaking down traditional labor/wage boundaries (e.g., SDR research in the US vs. global gig competition).
- "You're not judging it based on somebody in the Philippines versus somebody in India versus somebody in South America...all that is abstracted away." — Jason Calacanis [74:13]
-
TAO Investment Thesis:
- Jason: “TAO today is a 2–3 billion market cap. I believe that could become as big as Solana or Ethereum...if it works, it could be a 200x."
- Makes clear he is ready to lose his investment, but is participating to learn and place a calculated bet.
Notable Quotes & Memorable Moments
-
On the new founder skill-bar in the AI era:
- “Everybody became a superhero overnight. It was like somebody got the serum and everybody becomes a mutant and has crazy abilities.” — Jason Calacanis [09:58]
-
On diligence & culture:
- “We love a good hacker. We love somebody who can bend the will of the world to their vision...but don’t ever exaggerate reality.” — Jason Calacanis [22:18]
-
On crypto investment psychology:
- "I don't give investment advice, but what do you like to waste money on...I'm gonna buy $5,000 worth of tau. And now I'm watching it every day. Now I'm interested." — Jason Calacanis [85:01]
Key Timestamps
| Timestamp | Topic/Segment | |------------|-----------------------------------------------| | 04:09 | AI’s impact on early-stage startup formation | | 13:20 | Intro to Delve scandal with Ryan Madhavi | | 14:49 | Deep dive into allegations against Delve | | 19:36 | “Hustling” vs. “faking it”—founder/investor culture | | 22:09 | Investor diligence: red & pink flags | | 27:54 | How a quick demo could have exposed Delve | | 34:44 | Delve’s “donut” growth hacks & brand-washing | | 45:28 | Interview: Brick (AI energy optimization) | | 58:51 | Tao, decentralized AI/Web3/agentic future | | 61:33 | Guest: LeadPoet/Bittensor subnet, demo/tour | | 81:07 | Jason’s TAO investment thesis | | 85:01 | Comparing crypto bets to “cost of learning” |
Conclusion & Takeaways
- AI and automation keep lowering the technical/financial barriers for ambitious founders, but also raise their collective bar—founders need to continuously level up.
- Culture and incentives in the startup world often nudge toward “bending reality,” which can quickly cross into dangerous territory without strong governance and due diligence.
- In rapidly growing markets, investors and customers must demand more proof, product demos, and real customer validation, rather than relying on proxies or hype.
- Decentralized ecosystems (Web3/AI hybrids) are breeding new economic models that rewrite how tech, data, and human effort are valued, and are attacking price/cost structures across industries.
- The compliance scandal at Delve may be a preview of what happens when automation, hype, and lack of oversight collide—risking not just individual companies, but ecosystem-level trust.
For founders and investors alike, this episode is a masterclass in the realities—both inspiring and cautionary—of startup growth, trust, diligence, and what’s quickly becoming possible in the age of hyperautomation and decentralized innovation.
