Dan Gray
Research Lead at Odin
1 Oct 2026Super Intelligence | Dot’s Launch | Dario’s PR Team | Legora & H-Company CEO Join
20 Aug 2026Protein Binding | OpenAI Q2 Revenue | META Trial | TIME’s VC Rankings
Dan Gray: Time's VC Rankings & OpenAI Contrarian Take
Dan Gray joins from the Welsh countryside to critique Time's VC Power Rankings, whose methodology weights 50% on scale (capital raised, AUM, dry powder, deal volume) and predictably surfaces names like a16z and General Catalyst; he explains OrbiMed's incubation model and re-weights Ilya Strebulaev's approach by dollar-weighted efficiency, which lifts firms like SV Angel, Benchmark and USV. On the day's news he gives a contrarian take: OpenAI and Anthropic's reported revenue gap is partly accounting convention (~$10B of run rate), and since the two will converge over time, OpenAI at a lower IPO price may be the better long-term investment. He doubts frontier-lab economics, says LLMs aren't a path to AGI, and expects a correction after listing.
Thiago Silva: Prosus' 85-Company Robotics Report
Thiago Silva of Prosus presents the firm's new robotics report, based on visiting 85 companies across the US, China and Europe. He says robotics investment multiplied ~5x in the last year, humanoid prices have fallen from ~$150K to $50-60K and will reach $20-30K in two to three years, triggering a 'ChatGPT moment' for robotics and mass adoption — including a $400/month home robot that folds clothes. He covers automakers as the scaling playbook (Tesla pulled two models to build Optimus), China's government-backed push and Unitree's oversubscribed IPO versus Europe's regulatory hurdles, and the data bottleneck around hard-to-predict physics like deforming parcels.
Ruxandra Teslo Breaks Down Anthropic's Protein Results
Ruxandra Teslo, writer and editor at Works In Progress, explains the science: drugs target proteins, AlphaFold predicts protein structures, and models like David Baker's RF Diffusion (2023) already generate binders — Claude Science acts as a 'mini scientist' orchestrating these open-source tools with specialized tacit knowledge. She argues binder design is not the main bottleneck in drug discovery (finding which proteins to target is), so this is not an AlphaFold moment; the real significance is automating scientific pipelines. She notes the biology community is broadly excited about AI but disappointed the open-source models weren't credited, and she's skeptical all diseases will be cured in 10 years.
ChatGPT for Teens & Book Banter
After thanking Ruxandra and recapping the day's guests, the hosts banter about Progress magazine books before digging into OpenAI's new ChatGPT for Teens — a student-friendly experience automatically applied to under-18 accounts, with study mode, homework guidance, break reminders and stronger guardrails against emotional dependence. OpenAI uses an age-prediction model with ~2,000 signals, including chat content and times of day, to route teens into the restricted experience, and adults can appeal misclassifications. The hosts speculate on how the detection works and whether it can be gamed.
Claude & Grok Guess the Hosts' Ages + Meta's $1.4T Lawsuit
The hosts prompt Claude, ChatGPT and Super Grok to guess their ages from chat context. Claude puts one host in his late 20s to mid-30s when he's actually 24 and guesses 31 for another who just turned 32, reasoning partly from lowercase typing, while Sam's assistant refuses to guess and Max's Super Grok turns out to have no memory of him. Claude also surfaces news that four US states are suing Meta for up to $1.4 trillion over its alleged contribution to youth mental-health crises, in a trial that started Tuesday.
Ofir Ehrlich (Eon): Data Is the New Gold
Ofir Ehrlich introduces Eon, which finds, backs up, retains and governs enterprise data so it can be safely activated with AI tools. He points to Google buying $10 million of Spirit Airlines' data out of bankruptcy as evidence that data is now the most valuable asset, says most companies don't even know where their data is, and describes how tedious it is for data teams to locate, retrieve and activate it. He predicts labs' data appetite will spread to every company, argues data and people are the remaining moats as software stops being one, discusses possible acquisitions made purely for data, and says that as an angel investor in 70+ companies he's focused on AI infrastructure.
4 Feb 2026Moltbook🦞, SpaceX Buys xAI, Forbes 30u30, Granola Rebrand
Productivity Shocks & Investing in AI's Consequences
Dan frames AI step-changes (e.g., the nano banana image model, Genie world-generation) as productivity shocks that historically trigger IPO waves, and argues the remaining alpha is in the "consequences of AI" — like desalination powered by surplus energy from data-center overbuild. The hosts bring up memory stocks like SanDisk running up ~1,200% on AI demand.
Stagnation, Wasted Capital & the Crypto Hype Cycle
Dan laments the boom-bust venture cycle: cheap capital flows into slop (e-scooters, rapid delivery, NFTs) while energy and defense tech starve, and capital only chases crypto when Bitcoin is up. He also briefly plugs his reserves-paradox guide for emerging managers before wrapping up.
24 Oct 2025etn. Episode 4
Dan Gray: Europe's VC Gap Starts at First Check
Dan Gray, head of insights at startup valuation platform Equidam, walks through his post written a year after Mario Draghi's EU competitiveness report: the much-cited 'growth capital gap' is a symptom, since too few scale-ups earn late-stage money, and the real bottleneck is inception/first-check funding. He notes the best European scale-ups take US growth rounds from Sequoia or a16z, extracting value and exits from Europe, and contrasts a ~$16B US angel market with the EU's ~$1.2B. US early-stage funds also make 20K–100K pilot checks, while European LPs — not just GPs — stay too risk-averse to back that strategy.
Dan Gray's Fix: Move Government Capital to the Earliest Stages
Dan's prescription is reallocation rather than more money: research shows shifting existing government VC allocation to the earliest stages yields the same productivity gains as doubling it, and the British Business Bank's new emerging-manager program is a first step. He cites data that EIF-backed fund ones beat market exit rates by ~7% and create more jobs, while Horizon sends ~60% of tech funding to mid-tech firms and repeat corporate recipients. Early capital is generative and diversifies what gets funded, he argues, but requires thinking beyond a single election cycle.