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
Episode
73 min
Read time
3 min
Topics
Productivity, Relationships, Investing
AI-Generated Summary
Key Takeaways
- ✓Reliability Over Raw Intelligence: The core bottleneck in AI automation is not capability but variance — models perform at genius level roughly 95% of the time and fail catastrophically the remaining 5%. For background automation, that failure rate makes expected value calculations unworkable. Accounting, customer service, and drive-throughs keep failing not because AI lacks intelligence, but because inconsistent output cannot be trusted for high-stakes decisions without human oversight.
- ✓AI's Bitterest Lesson — Data Hierarchy: In order of training importance: defining the right task comes first, then data, then compute, then algorithms. Compute scaling alone plateaued around GPT-4.5, which remains the largest model by token cost yet failed commercially. Data is many times more impactful than compute, but the correct data only becomes definable once the right task is identified — making task definition the true unlock for model improvement.
- ✓Benchmark Manipulation by Tier: Top-tier labs like OpenAI and Anthropic have little incentive to benchmark-max because they genuinely compete for first place. Second-tier labs — including several Chinese labs and American open-source contenders — face internal pressure to appear competitive and systematically inflate benchmark results. Decision model benchmarks are particularly unreliable because no consensus exists on what good performance even means in that category.
- ✓Why Enterprise Software Hasn't Changed: Enterprises rationally avoid embedding AI into core application logic because unreliable models cannot be trusted for company-level decisions. Instead, AI gets attached to user-facing surfaces where errors shift liability onto users. The result is unchanged SaaS architecture with AI bolted on the side. Real transformation requires reliability high enough to automate multi-week background tasks, not just assist humans with incremental efficiency gains of 20%.
- ✓Jevons Paradox as Growth Model: Jev's path to scaled revenue relies on triggering a mega-Jevons paradox — making AI cheap and reliable enough that projects previously impossible become viable, expanding total AI budget rather than just reducing per-unit cost. Almeida argues current automation across all major coding agents and decision models collectively sits near zero percent of addressable automation work, meaning the market being built essentially does not yet exist.
What It Covers
Diogo Almeida, founder of TypeSafe and creator of Jev — a decision model used by one-third of Fortune 500 companies and valued at $7.5B — explains why reliability of AI intelligence, not compute or benchmarks, determines real-world automation success, and why current AI deployment remains fundamentally broken across enterprise software.
Key Questions Answered
- •Reliability Over Raw Intelligence: The core bottleneck in AI automation is not capability but variance — models perform at genius level roughly 95% of the time and fail catastrophically the remaining 5%. For background automation, that failure rate makes expected value calculations unworkable. Accounting, customer service, and drive-throughs keep failing not because AI lacks intelligence, but because inconsistent output cannot be trusted for high-stakes decisions without human oversight.
- •AI's Bitterest Lesson — Data Hierarchy: In order of training importance: defining the right task comes first, then data, then compute, then algorithms. Compute scaling alone plateaued around GPT-4.5, which remains the largest model by token cost yet failed commercially. Data is many times more impactful than compute, but the correct data only becomes definable once the right task is identified — making task definition the true unlock for model improvement.
- •Benchmark Manipulation by Tier: Top-tier labs like OpenAI and Anthropic have little incentive to benchmark-max because they genuinely compete for first place. Second-tier labs — including several Chinese labs and American open-source contenders — face internal pressure to appear competitive and systematically inflate benchmark results. Decision model benchmarks are particularly unreliable because no consensus exists on what good performance even means in that category.
- •Why Enterprise Software Hasn't Changed: Enterprises rationally avoid embedding AI into core application logic because unreliable models cannot be trusted for company-level decisions. Instead, AI gets attached to user-facing surfaces where errors shift liability onto users. The result is unchanged SaaS architecture with AI bolted on the side. Real transformation requires reliability high enough to automate multi-week background tasks, not just assist humans with incremental efficiency gains of 20%.
- •Jevons Paradox as Growth Model: Jev's path to scaled revenue relies on triggering a mega-Jevons paradox — making AI cheap and reliable enough that projects previously impossible become viable, expanding total AI budget rather than just reducing per-unit cost. Almeida argues current automation across all major coding agents and decision models collectively sits near zero percent of addressable automation work, meaning the market being built essentially does not yet exist.
- •Fundraising Without Valuation-Maximizing: TypeSafe raised at a $7.5B valuation without a pitch deck, running entirely on inbound interest. The team explicitly rejected cuck-tranche structures and maximum valuation as goals, instead optimizing for strategic allies capable of helping navigate competition with the largest technology companies. Andreessen Horowitz was chosen for philosophical alignment; Sequoia was chosen specifically for partner Sonya Wang's distinct strategic thinking, not the fund's brand.
Notable Moment
Almeida revealed he almost abandoned the TypeSafe concept before founding it because the direction seemed so obviously correct that he assumed Anthropic was already 80% finished building it. He only proceeded after concluding the field had genuinely missed the insight — a bet that produced $100M in revenue within seven days of launch.
Episode Transcript
Almost everyone who's paying attention to AI knows reliability is the absolute largest bottleneck. You get Einstein, like, maybe 95% of the time and mister Bean, the other 5%. Data matters a lot more than compute. I don't want to be called a NeoLab because from my sense, most NeoLabs are incredibly, incredibly wasteful. Anthropic models are just the best even if the Chinese models match it in benchmarks. It feels like there's no real competition. Right now. I truly believe that we're gonna, like, mega Jevan's paradox. AI is so clearly awesome, yet software itself hasn't changed. Everyone has just been in this, like, mono model, like, horse blinders for so fucking long. I believe that AI has been a bubble for a while, so I would like to keep planning for an eventuality of bubble burst. What we need is allies to fight the largest companies in the world. We don't need the money or the valuation. This is 20 VC with me, Harry Stabenowitz. Now joining me in the hot seat today is one of the best founders of our time, one of the hottest companies, Diogo Almeida. Diogo is the founder of TypeSafe, the maker of JEV. JEV is used by a third of the Fortune five hundred. Pretty incredible. Yep. It gets better. Sequoia's Pat Grady says JEV went from zero to 100,000,000 in revenue in just seven days. And when you're that fast and that good, it comes with some benefits. They've been valued at $7,500,000,000, and they had their round led by Martin Casado at Andreessen, and Sequoia also invested. Today, we get into the future of models with Diogo. This is an incredible discussion. But before we dive into the show today, Blitzy, the autonomous software development platform that's built for enterprise scale, is proud to sponsor today's show. Enterprise code bases have grown beyond human comprehension. Tens of millions of lines, decades of tech debt, complexity existing tools cannot fathom. Blitzy starts by understanding. Thousands of agents reverse engineer your code base into a proprietary knowledge graph that maps architecture, dependencies, and business logic. That graph surfaces every CVE and zero day hiding in your code before anyone hands you a ticket. Grounded in that context, Blitzy then executes entire epics autonomously, modernize off legacy systems, ship new features five x faster, or batch remediate CVEs at scale surfaced by the platform. Blitzy not only generates code, it validates an end to end test every single line, and that's why Fortune 500 enterprises are shipping enterprise grade code five times faster. Go to blitzy.com/20vc to learn more. While Blitzy helps you build faster, Flex helps you run your finances smarter. Business owners run their whole financial life on Flex. One platform from business revenue to their personal spend. Float every purchase for sixty days. Tap capital that grows with your revenue and pay vendors in a 170 countries across 32 currencies, plus the whole back office, bills, expenses, accounting, all in …
Get the full transcript (14,459 words) + summary by email — free
One-time email with the complete transcript and AI summary of this episode. No account needed.
One email, no spam. We’ll also show you what SignalCast does.
You just read a 3-minute summary of a 70-minute episode.
Get 20VC (20 Minute VC) summarized like this every Monday — plus up to 2 more podcasts, free.
Pick Your Podcasts — FreeKeep Reading
More from 20VC (20 Minute VC)
20VC: Taking on NVIDIA: Fractile CEO on the Future of Chips | What Everyone Gets Wrong About Capex Spending | The Winners and Losers in Neo Clouds with Walter Goodwin
Nov 14 · 75 min
a16z Podcast
AI Can Write Code. Why Isn’t Software Better?
Sep 28
More from 20VC (20 Minute VC)
20VC: Jensen Huang Investigation: What Happens Now | Sequoia Leave London, Close Scout Program and End Evergreen Fund | NASDAQ Loses $2.5TRN in 24 hours | Anthropic targets $1.4TRN IPO valuation
Nov 12 · 66 min
The AI Breakdown
Why a New Class of AI “Judgment Models” Could Have Big Business Implications
Sep 16
More from 20VC (20 Minute VC)
We summarize every new episode. Want them in your inbox?
20VC: Taking on NVIDIA: Fractile CEO on the Future of Chips | What Everyone Gets Wrong About Capex Spending | The Winners and Losers in Neo Clouds with Walter Goodwin
20VC: Jensen Huang Investigation: What Happens Now | Sequoia Leave London, Close Scout Program and End Evergreen Fund | NASDAQ Loses $2.5TRN in 24 hours | Anthropic targets $1.4TRN IPO valuation
20VC: Mistral CEO on Why Energy Is Not the Bottleneck to AI | Why OpenAI and Anthropic Will Come for Your Business | The Future of the Inference Market | Are Chinese Open-Source Models Truly Safe?
20VC: From $8BN to $19BN: Inside WashU's Endowment Fund | SpaceX Made Up 20% of Our AUM | Why Europe and China Has So Much Opportunity | The Reckoning That is Coming for AI Companies with Scott Wilson
20VC: OpenAI Cuts Valuation to $1.25TRN | Public Market Meltdown: GOOG Loses $500BN in 12 Hours | DoorDash Buys Instacart for $17BN | Kalshi Raises $1BN and Cognition Eyes $80BN Valuation
Similar Episodes
Related episodes from other podcasts
a16z Podcast
Sep 28
AI Can Write Code. Why Isn’t Software Better?
The AI Breakdown
Sep 16
Why a New Class of AI “Judgment Models” Could Have Big Business Implications
Invest Like the Best with Patrick O'Shaughnessy
Sep 1
Sarah Guo - What the 250 People Building AI Believe - [Invest Like the Best, EP.489]
Latent Space
Aug 21
Simulation: the new Scaling Law — Joon Sung Park, Simile AI
This Week in Startups
Aug 21
Open source is going to win it all: Harvey proves it | E2328
Explore Related Topics
This podcast is featured in Best Investing Podcasts (2026) — ranked and reviewed with AI summaries.
Read this week's Investing & Markets Podcast Insights — cross-podcast analysis updated weekly.
You're clearly into 20VC (20 Minute VC).
Every Monday, we deliver AI summaries of the latest episodes from 20VC (20 Minute VC) and 192+ other podcasts. Free for one show.
Start My Monday DigestNo credit card · Unsubscribe anytime