SHOW DIRECTORY
The a16z Show
The venture firm’s view of startups, AI, software, markets, company building, and the technological shifts creating new categories.
BIDCLUB DESCRIPTION
What It Takes to Run a Personal Assistant Around the Clock
Ken LiuMartin CasadoJoel de la Garza
Second aims to be a portable store of personal context that specialized agents can query, with background monitoring linking events to priorities without prompts.Ken Liu estimates ten always-on agents could cost $1,000 daily at “Opus 5.5” pricing versus $60 with DeepSeek Flash, making model routing central to the economics.Local data control and confidential GPU inference shape the security design, but zero data retention remains a policy promise and prompt injection cannot be definitively solved.
What History Tells Us About the AI Revolution | Tyler Cowen & Alex Imas
Tyler CowenAlex ImasErik Torenberg
AI capabilities are advancing quickly in areas with few human bottlenecks, but Tyler Cowen and Alex Imas see no evidence yet for sustained double-digit growth.Ramp data show token use rising exponentially while revenue stays flat as price competition intensifies, challenging the assumption that usage growth will support frontier-lab forecasts.AI-driven productivity may not translate into durable rents for AI capital owners; value could accrue to scarce energy, land, or human work.
How to Defend Against AI Hackers, with Wiz’s Ami Luttwak
Ami LuttwakZane LackeySarah Wang
AI can complete attacks that once took elite hackers weeks or months in roughly 20 minutes, while open-source models may trail the frontier by no more than three months in vulnerability discovery.Patching lags: about 50% of Wiz customers patched the PHP vulnerability within two or three days, while older versions lack patches and Linux fixes can take weeks or months to reach users.
How AI Could Bring Halo and Other Classic Games Back to Life
AI agents that operate development tools and correct their output could shorten prototyping; Anish Acharya tentatively imagines 5–10-year cycles falling to 2–5 years.Lower costs could let studios test more ideas, but human taste still shapes the final 20% of a game.Community Halo projects show preservation potential, while how IP holders capture value remains unresolved.
Mike Volpi on Why AI Infrastructure Will Outlast the Hype Cycle
Mike VolpiRaghu RaghuramGuido Appenzeller
AI capacity may remain useful even if financially weak neoclouds fail, as AI providers’ revenue growth has outpaced the earlier cycle.GPU scarcity is keeping older H100s in use and rents elevated, but token revenue per watt could make them uneconomic once Vera Rubin supply catches up.Silicon constraints may persist for years, while Cisco’s order collapse shows how shortage-driven demand can reverse sharply when supply commitments outlast actual need.
Where Startups Can Win Beyond OpenAI and Anthropic | HubSpot’s Dharmesh Shah
Shah’s test is whether an opening survives “yet”; structural or incentive constraints can protect it from OpenAI and Anthropic.OpenRouter and task routing reduce model dependence; business apps may retain value beyond the interface.HubSpot’s solo-entrepreneur CRM tests demand amid roughly 5.6 million new U.S. companies in 2025; over 70% are not expected to hire within two years.
How Salesforce Chooses the AI Companies It Wants to Buy
Salesforce pairs scarce AI capabilities with 15,000 sales reps, though six-to-nine-month integrations keep build-versus-buy speed in question.Fin’s 14-day internal deployment, over-40% share of help.salesforce.com traffic and 65% human-free resolution illustrate outcome pricing as customers may need fewer seats.Somorjai says security can veto deals, while valuations above $10 billion leave few buyers and the timing of a reckoning remains unknown.
How AI Is Turning Figma’s Finance Team Into Builders
Praveer MelwaniSeema AmbleIvan Makarov
Figma’s finance team is building beyond analysis, using AI for collections and product replicas that test technical difficulty and moat depth during diligence.Melwani says headcount could fall a year from now if complexity stayed flat, but new products and acquisitions keep adding work as efficiency improves.Scaling local tools into shared infrastructure and producing audit-ready outputs remain key tests.
The New AI Agents That Shop, Plan, and Gossip
David PawlanAnish AcharyaJosh Elman
Consumer agents may span general, work, multiplayer, and travel, but trust looms: 22% of roughly 3,600 group-chat messages cited it as the main fear.Shopify-enabled direct discovery could pressure aggregators, while agents may favor “clean APIs with great docs” over better products.Execution still differentiates at scale, though David Pawlan thinks rivals might cross that moat in “a year or so.”
Andrew Huberman on Focus, Energy, and Performing at Your Best
Huberman says founders should protect their peak work windows—9:30 a.m.–12:30 p.m. and roughly 4:30–5 p.m. through 7 p.m.—and exercise before 9 a.m., which he says improves his energy and focus.Sleep and waking look like earlier neurotechnology opportunities because their target states are clearer, while desired brain-activity patterns for specific work tasks remain unknown.









