全部/科技/实时热榜

Tomer Tunguz · 实时热榜

56分钟前更新

01
AI Worldviews
The Economist scored 25 frontier AI models on the World Values Survey. Lab of origin is a weaker predictor than training & alignment choices : Gemini & Qwen are neighbors, GPT-4o & DeepSeek R1 are near-twins, & DeepSeek R1 & DeepSeek V4 Flash are strangers. Worldview is invisible in code generation. In business analysis, forecasts, hiring, & policy work, it is a live input.
02
Search
Search the tomtunguz.com archive of essays on AI, SaaS, startups, and venture capital.
03
Most AI Work Can Wait
Prioritize routing over model choice. Most AI work runs on cheap local models.
04
The CIO's Choices are Clear in 2026
The CIO's choices are clear in 2026. Across 87 public SaaS & platform companies, only Infrastructure & Dev Tools (+68.5% 1Y) & Security (+17.6% 1Y) are positive; the other three sectors are down. Category, not growth, separates winners from losers : the market buys the AI stack & sells the seat-priced application layer.
05
When AI Costs More Than the Engineer
Anthropic spends 2.3x its payroll on compute — $515k per engineer per year at today's $224k fully-loaded salary. The top 1% of software companies spend $89k, the median $137. Three 2029 scenarios bracket how that gap closes.
06
What If There Is No Moat Yet?
At the application layer, moats are lagging, earned through scale & brand. At infrastructure, capital intensity demands a leading moat at founding.
07
Full Sail on Asynchronous Inference
The inference market is the largest in software. AI workloads are shifting from synchronous chat to asynchronous, multi-turn agents that run for hours. Sail Research builds fleet-aware orchestration to maximize throughput per dollar of inference spend.
08
Defending Against AI-Powered Attackers
Sunil Agrawal, CISO at Glean, joins Office Hours to discuss what security readiness looks like when attackers use frontier models for reconnaissance, phishing, deepfakes, and exploit generation.
09
The Quietest Part of Startupland isn't so Quiet
Stablecoins hold $165b in US T-bills, more than China or Switzerland. Hyperliquid generates $610m annualized fees. Gold, oil, & stocks trade 24/7 on crypto perpetual markets.
10
So You Want to Sell Inference
Reselling inference at cost is a zero-margin business. The margin question is cost-plus vs. value-based pricing. Optimization is a cost lever under either, with distillation defensible for a while. BYOK breaks cost-plus & spares value-based.