20VC: Jev Creator on Why Data Is More Important Than Compute | Will US Open-Source Models Catch Chinese | The Cost of Intelligence and Tokens: What No One Sees But Needs to Know with Diogo Almeida
20VC: Jev Creator on Why Data Is More Important Than Compute | Will US Open-Source Models Catch Chinese | The Cost of Intelligence and Tokens: What No One Sees But Needs to Know with Diogo Almeida
Podcast1 hr 13 min
Listen to Episode
Note: AI-generated summary based on third-party content. Not financial advice. Read more.
Quick Insights
  • Treat Jev as a high-risk private-company opportunity, not a publicly traded stock; its reported $7.5 billion valuation and rapid revenue claims are unverified, and automation adoption remains uncertain.
  • Be cautious on AI infrastructure and neoclouds: the discussion flags a potential demand shock if AI fails to meet expectations, but gives no specific short trade or price target.
  • For public software investments such as Salesforce and Intuit, look for proof of reliable, end-to-end workflow automation rather than superficial AI features; no buy recommendation or valuation target is provided.
Detailed Analysis

TypeSafe / Jev (Private company)

  • Jev is described as an AI model product focused on reliability of intelligence—making AI dependable enough to automate work in the background, rather than merely assist people.
  • The founder says Jev is used by a third of the Fortune 500 and claims it went from zero to $100 million in revenue in seven days. The transcript does not clarify the revenue measure or timeframe behind that claim.
  • The company is said to have been valued at $7.5 billion in a round involving Andreessen Horowitz and Sequoia. The founder says the company is profitable, did not need to raise urgently, and prioritized strategic investors over maximizing valuation or funding.
  • The founder’s central thesis is that AI progress depends more on choosing the right task and building suitable data than on simply adding compute. He argues that reliability, not just speed or cost, is what will unlock automation.

Takeaways

  • Jev represents a high-risk, private AI investment opportunity, not a publicly traded stock. The valuation and growth claims are reported in the conversation and are not independently verified by the transcript.
  • The investment thesis depends on Jev turning improved reliability into durable, large-scale automation. The founder acknowledges that real automation remains near zero across the broader field, so adoption and execution are still significant uncertainties.
  • For public-market investors, the discussion points to reliability and practical deployment as themes to watch—not to a specific buy recommendation or price target.

AI Data Providers (Surge and McCaw)

  • The discussion considers whether data providers such as Surge and McCaw can become attractive businesses as AI labs spend on data.
  • The guest sees potential in a large market for data, including future demand from companies developing their own AI capabilities. He cautions, however, that the amount AI companies will ultimately spend on data as sustainable businesses is uncertain.
  • He emphasizes that useful data is task-specific and often must be created or shaped for a particular goal; simply collecting existing corporate data may not be enough.

Takeaways

  • AI data services may benefit if demand for task-specific training and evaluation data grows, but the conversation offers no firm revenue outlook or investment recommendation.
  • Assess whether a provider can produce data that measurably improves models for important tasks, rather than relying on broad claims about data demand.

AI Model Developers and Open-Source Models

  • The guest argues that benchmark results can be misleading and that models should be judged by how reliably they perform useful work.
  • He views Anthropic as the strongest model provider in the current landscape, even when competitors match it on benchmarks. He says the large labs, including OpenAI and Anthropic, have less need to rely on benchmark manipulation than second-tier competitors.
  • He is skeptical that U.S. open-source models will catch up with Chinese models in raw capability, partly because U.S. developers may face higher ethical and legal constraints. He also says a capability gap might matter less if model progress plateaus.
  • He would not use a Chinese model “vanilla” in a coding agent, citing security concerns, while noting that other safeguards and model adjustments could address some risks.
  • The guest characterizes parts of the AI investment surge as wasteful and says he believes AI has been in a bubble for some time.

Takeaways

  • The discussion favors evaluating model companies on reliability, real-world usefulness, and efficiency, rather than headline benchmark scores alone.
  • Model development carries substantial risks: the guest raises the possibility of an AI investment bubble, uncertain returns on spending, and security concerns around deploying some models without safeguards.
  • No specific public stock, price target, or recommendation is given. The named model developers are not presented as direct investment recommendations.

AI Infrastructure and Neoclouds

  • The guest is cautious about neoclouds and the broader AI hardware buildout. He says he is planning for the possibility of a bubble burst caused by a demand shock, rather than a shortage of supply.
  • He suggests demand could weaken if companies discover that AI cannot deliver what they expected. He also notes that spending may be partly driven by fear of missing out.
  • He says Jev’s efficiency could help it avoid being heavily dependent on the GPU spending associated with the standard subsidized growth playbook.

Takeaways

  • AI infrastructure and neocloud investments may be vulnerable if AI demand falls short of expectations, even if the technology continues to improve.
  • The transcript supports a cautious approach to companies whose outlook depends on sustained, rapid growth in AI compute demand. It provides no specific infrastructure stock or price target.

Enterprise Software and SaaS

  • The guest rejects the idea that a broad “SaaSpocalypse” is inevitable. He says established software companies may be well positioned because they understand customer workflows and can invest in the engineering needed to automate them.
  • He cites Salesforce and Intuit as examples of incumbents that may have relevant workflow knowledge. He argues that their potential advantage is not simply distribution, but proximity to customer problems.
  • He expects AI to be more transformative when it automates work reliably, rather than being added as a side feature or a co-pilot that leaves users responsible for checking its output.
  • He says companies’ proprietary “intelligence” may come more from the software built around models than from training their own models.

Takeaways

  • The discussion suggests that established software firms with deep workflow knowledge could be positioned to benefit from AI—but only if they build useful automation, not just add superficial AI features.
  • A practical way to assess software companies is to look for evidence of reliable end-to-end task automation, while recognizing that the guest says this capability is still largely unrealized.
  • The conversation does not provide specific stock recommendations or forecasts for Salesforce, Intuit, or the broader SaaS sector.

AI Applications and Coding Agents

  • The conversation mentions Cognition, Harvey, Cursor, Factory, and Modal as examples in the coding-agent and AI application landscape. These are discussed as companies or products, not as specific investment recommendations.
  • The guest believes coding agents have improved substantially, but argues that many AI products still assist workers rather than automate entire tasks.
  • He is skeptical of claims that application companies have a durable moat solely because they train their own models. In his view, deep software engineering around real workflows may be a more plausible source of differentiation.
  • He also warns that AI hype has moved in cycles, with past waves of AI companies rising and then fading.

Takeaways

  • For AI application companies, the discussion favors looking for measurable workflow automation and sustained customer value, rather than relying only on claims about proprietary models or benchmark performance.
  • The guest’s view is that AI applications may have considerable upside if they achieve reliable automation, but the transcript also highlights execution risk and the possibility that current enthusiasm exceeds what the technology can deliver.
  • No specific company is recommended for investment.

Robotics

  • The guest calls robotics the most overhyped category in his view, not because it lacks promise, but because he believes reliable AI is still far from the level robotics requires.

Takeaways

  • The discussion signals caution toward robotics businesses whose investment case depends on near-term, dependable AI performance in physical environments.
  • This is the guest’s opinion, not a specific forecast or recommendation.

Sequoia and Andreessen Horowitz (Private venture-capital firms)

  • Sequoia and Andreessen Horowitz (a16z) are identified as investors in Jev’s funding round.
  • The Jev founder says he selected them for strategic support and alignment, particularly valuing a16z’s technology-focused outlook and Sequoia investor Sonali Wong’s strategic perspective.

Takeaways

  • The firms’ participation indicates that prominent venture investors backed Jev, but it does not establish that the investment will succeed or provide a way for public-market investors to invest directly.
  • The transcript provides no performance figures or investment recommendation for either venture firm.

Cryptocurrencies

  • No cryptocurrencies or crypto investment opportunities are mentioned in the transcript.
Ask about this postAnswers are grounded in this post's content.
Episode Description
Diogo Almeida is the Co-Founder and CEO of TypeSafe AI, the company behind Jev. Jev is an AI model that lets software make decisions automatically and tells developers how confident it is in each call. According to Sequoia's Pat Grady, it went from zero to $100M in revenue in seven days. TypeSafe's latest round valued the company at $7.5BN, with Martin Casado at Andreessen Horowitz leading and Sequoia also investing. Before TypeSafe, Diogo was a researcher at OpenAI, where he helped invent ChatGPT.  AGENDA:  00:00 Why reliability is the biggest bottleneck to AI 10:26 When does doubling compute stop making models smarter? 12:06 If data matters more than compute, which data actually makes models better? 22:00 What stops OpenAI from copying your product the moment it works? 37:53 Are Harvey and Legora just wrappers around frontier models? 40:05 How much smarter is AI actually getting per token? 43:26 Which AI lab makes the most misleading benchmark claims? 48:19 Will American open-source models catch up with China's? 50:01 If demand for GPUs keeps growing, what bursts the AI bubble? 57:39: How we raised $750M in a week with no deck?  1:07:03 Why Typesafe does not hire from OpenAI?
About The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch

The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch

By Harry Stebbings

The Twenty Minute VC (20VC) interviews the world's greatest venture capitalists with prior guests including Sequoia's Doug Leone and Benchmark's Bill Gurley. Once per week, 20VC Host, Harry Stebbings is also joined by one of the great founders of our time with prior founder episodes from Spotify's Daniel Ek, Linkedin's Reid Hoffman, and Snowflake's Frank Slootman. If you would like to see more of The Twenty Minute VC (20VC), head to www.20vc.com for more information on the podcast, show notes, resources and more.