This Week in Startups E2030: Election Security, Skift's Rafat Ali, and Anthropic Unleashes PC AI

Date: October 23, 2024
Host: Jason Calacanis
Co-host: Alex Wilhelm
Guests: Hans von Spakovsky (Heritage Foundation), Rafat Ali (Skift)

Episode Overview

This episode dives deep into current themes shaping the startup and tech landscape, focusing on three core topics:

1. Election Security in the US, including insights from the Heritage Foundation’s election fraud database
2. The changing travel industry and tech M&A, with Skift founder Rafat Ali
3. Cutting-edge AI advances, especially Anthropic’s new capabilities to control computers via AI

The show maintains a tone of critical inquiry, aiming to challenge guests on their assertions while seeking actionable insight for founders, operators, and curious listeners.

Key Segment Summaries & Insights

1. Election Security & Fraud in the US

Setting the Context (00:00–03:36)

• The episode opens with concerns over past contested elections and a preview of the guests to come.
• Jason frames the discussion: technology’s role in election security and how real the threat of fraud actually is.

Hans von Spakovsky on Heritage Foundation’s Election Fraud Database (04:10–11:19)

• Purpose of the Database: Heritage sought to fill a gap with a nonpartisan resource of “proven cases” of election fraud—i.e., only after conviction or official government finding.

  "It's only proven cases... convicted in a court of law, a judge orders a new election because of fraud, or there's an official finding by a government agency." (Hans, 04:24)

• Database Stats:

  • About 1,600 cases, spanning back to the early 1980s, all levels (local to federal).
  • Not comprehensive; most are local, many cases unreported or unprosecuted.

  "We're approaching 1,600 cases... keep in mind, this is not a comprehensive list." (Hans, 07:01)

• Interpreting the Numbers: The number is tiny compared to total votes—but Hans cautions that most fraud is never prosecuted or even referred.

  "I don't think we have massive fraud... but I do think we have enough of it that we should be concerned..." (Hans, 09:35)

Technology & Solutions for Election Integrity (12:48–17:42)

• Heritage also runs an Election Integrity Scorecard: 50 criteria for best practices per state (ID, up-to-date voter rolls, absentee ballot protocols, etc.)

• Top score among states was 82/100 initially, now up to ~90.

• Technology’s lag in state governments is significant, due to funding and competitive pay for IT talent.

  "State governments are often far behind... particularly private industry, when it comes to their technology." (Hans, 16:13)

• Hans supports requiring free voter ID for all and monthly (ideally real-time) syncing of state databases.

• Calls for states to link DMV and public assistance databases to voter rolls.

On 2020 and Adjudication of Election Challenges (17:53–22:30)

• Jason asks directly about “who won” and whether the adjudication process was sufficient.

  "What should have happened, unfortunately didn't, is most of the lawsuits... never got to the substantive hearing stage." (Hans, 19:07)

• Hans repeatedly asserts that the court process should have been more robust, with claims litigated on the merits, but indirectly concedes that Biden won.

  "Biden won. He was declared the winner... there's no point in arguing about it today." (Hans, 20:36)

• Jason presses for clarity, which Hans avoids (“the claims that were made were never resolved the way they should have been”), reflecting broader partisan dance around the issue.

Penalties and Prevalence of Election Fraud (27:32–30:06)

• Penalties for fraud depend on prosecution level—some severe (jail), some as light as a fine.
• Hans recounts cases from recent years, e.g., bribes for ballot box stuffing, and non-citizen voting with only light consequences.

Panel Reflection on Election Security (33:04–37:35)

• Jason and Alex reflect that, despite highly publicized fears and scattered individual cases, the scale of fraud capable of swinging a national election is vanishingly small.

  "The idea that we could actually swing it one way or the other... feels like the numbers don't matter to me of there being systematic election fraud..." (Jason, 33:04)

• Alex calls out Heritage for sometimes fearmongering with "in camera footage" of minority voters, expressing a need for better, less inflammatory public discussion.

2. The Travel Industry, Uber-Expedia Rumors, and Skift’s Trajectory

Rafat Ali on Uber-Expedia (41:57–47:22)

• Travel booking sites face “infrequent usage” (2–4 times/year on average user), creating high customer acquisition costs.

• Uber’s 150M user base presents an upsell opportunity for travel bookings; Expedia is struggling and stock is languishing.

• Integration challenge: most travel bookings (especially in the West) are done on desktop, rides/orders on mobile, which complicates super app ambitions.

  "There's an overemphasis on assuming that Expedia and everything else in online travel is a huge brand that people have loyalties to. Most of their customers come through online acquisition channels—Google." (Rafat, 45:12)

• No real loyalty—travel shoppers price-comparison search across multiple platforms.

• Uber Travel initiatives in UK and India show higher rebooking rates for those who try multimodal bookings.

Loyalty Programs, Marketplaces, and Asset Light Models (48:58–56:31)

• Travel industry is “addicted” to loyalty programs—miles & points drive significant revenue and user stickiness for airlines and hotels.
• Uber 1 is gaining traction as a subscription model, but similar attempts in travel are still niche.
• VRBO (Expedia’s home rental arm) has untapped potential, especially post-pandemic, but now struggles in city markets.
• Mentor relationship between Dara Khosrowshahi (Uber CEO) and Barry Diller (Expedia chair) may grease the wheels for a deal.

AI, Industry Integration, and Future of Traveltech (57:00–62:40)

• AI could transform travel’s “tyranny of the search box,” enabling better user experiences and backend links between airlines, hotels, and transport.

• Skift’s own AI “Ask Skift” engine (SEC filings, B2B data) gets 1K–2K queries/day; not a hockey stick but steady.

• Licensing and indexing of Skift’s content by major AI companies is an ongoing issue—most powerful publications are seeing their content scraped and trained.

  “You have an unbiased source that’s high quality—man, that’s super valuable.” (Jason, 60:26)

Remote Work, Global Teams, and the Media Business (68:16–70:56)

• Skift has evolved to a 100-person fully remote, global workforce. The cost base is now much lower—and team loyalty is higher, especially outside the US.
• Niche B2B subscriptions (esp. with “high quality human content”) and corporate sales are the growth model for sustainable media businesses.

3. NEW AI CAPABILITIES: Anthropic & Beyond

Anthropic’s “Computer Use API” Demo (75:47–78:06)

• Anthropic released a feature to allow Claude (its AI model) to take over browser windows (and potentially full computers) to automate tasks.

• This enables agents to: scrape data, fill forms, book items, interact with services, etc., with natural language commands, which broadens the scope and ease of automation far beyond traditional RPA (Robotic Process Automation).

• Jason and Alex discuss both the power and the risks (e.g., automating scams, fake accounts).

  “This is now letting your language model run your browser. What could go wrong?” (Jason, 77:26)

• Potential for misuse is high, but so is the upside if properly harnessed.

Notable Quotes & Moments

Election Security

• Hans on structural weaknesses:

  “Most of the cases... were dismissed on procedural grounds... I’m not saying that the claims were legitimate. What I’m saying is the court process should have been worked to the end so that the questions would have been resolved.” (Hans, 00:16)

• Alex, challenging on process and public service pay:

  “When I heard that we were going to bring you on the show, I did not think that the guest from the Heritage foundation was going to be advocating for increased public servant salaries.” (Alex, 17:42)

• Hans conceding on the 2020 outcome:

  "Biden won. He was declared the winner... there's no point in arguing about it today." (Hans, 20:36)

• Jason, reflecting:

  “Even though a person did not want to concede, the democratic institution... allowed the transfer of power, even with January 6.” (Jason, 34:39)

Industry & Travel

• Rafat on Expedia and Uber:

  "There's just not loyalty in online travel... [customers] look at tens of sites before they actually book." (Rafat, 46:08)

• On AI integrating the travel industry:

  "The tyranny of the search box... is essentially the same in 2024 as it was in 1999. That just seems mind-blowing." (Rafat, 57:00)

• Remote, global work:

  “Having a mix of nationalities for a global industry like ours and having a diversified cost base... has outweighed the cons of not being in an office.” (Rafat, 69:10)

AI

• Alex, on Anthropic’s potential:

  "Offering this up as an API... to allow, I'm hoping, homebrew development with very modern AI models. That's so cool. I can't wait to see people dream up..." (Alex, 77:32)

• Jason, on risks:

  “You could have it... create a thousand accounts... DM people... and then ask them to send you, you know, a thousand dollars as booking fee to be on the podcast... you can think about the shenanigans this could do.” (Jason, 78:06)

Timestamps of Key Segments

• Election Security & Heritage Foundation:

  • 04:10–11:19: Building and maintaining the election fraud database
  • 12:48–17:42: Tech solutions and “integrity scorecard”
  • 17:53–22:30: 2020 election process and direct “who won” questions
  • 27:32–30:06: Penalties for fraud and case examples

• Travel Industry Deep Dive:

  • 41:57–47:22: Uber-Expedia pros, cons, and tech hurdles
  • 57:00–62:40: The role of AI and seamless travel
  • 68:16–70:56: Remote work and the business model for media

• AI & Automation News:

  • 75:47–78:06: Anthropic’s “Computer Use API” demo and implications

Conclusion/Takeaways

• US elections are far more secure than public discourse suggests, with proven fraud cases nearly statistically insignificant at the national level—but the system is still patchwork, with room for procedural and technical improvement.
• The future of travel and marketplaces likely lies with the integration of user data, loyalty, and AI—in which Uber, if it makes a bold acquisition or tech partnership, could dramatically shift the market.
• Anthropic’s API opening up browser and PC control to LLMs is a huge leap functionally—and a potential Pandora’s box if safety controls aren’t prioritized.
• Media businesses that are niche, B2B, and built on direct relationships (subscriptions, quality data, active LinkedIn presence) are thriving amid chaos in consumer tech/media.
• Critical, non-tribal conversations are still possible, and even necessary, with partisan guests—especially when hosts refuse to let the evasive “party-line” go unchallenged.

Hosts’ Closing Reflections:
Jason and Alex emphasize that both technology and society require constant vigilance, data-driven analysis, and open conversations—even (and especially) with those outside one’s own bubble. The episode closes with a look ahead at more coverage of AI and startup trends.

End of Summary