The Week AI Broke the Market: Models Building Themselves, SaaS De Rating, Musk in Space скачать в хорошем качестве

The Week AI Broke the Market: Models Building Themselves, SaaS De Rating, Musk in Space

01:22 — AI models: two flagships drop 20 minutes apart Anthropic ships Opus 4.6; OpenAI replies with GPT 5.3. Key tension: best model vs stickiest workflow (tooling + habits = raw benchmarks). China keeps coming: Kimi K2.5, Qwen3 Max — strong performance at lower cost, plus “swarm”/multi-agent vibes. 07:02 — Recursive AI + security flex OpenAI: “GPT 5.3 helped build itself” (debugging training pipeline). Anthropic: claims model found 500+ serious open-source security issues → “bots find bugs better than eyeballs.” 11:06 — Alphabet CapEx shock Alphabet expected $180B CapEx in 2026 → market flinches despite earnings beat. Take: hyperscalers signalling capacity constraint and “you ain’t seen nothing yet.” Debate: monster spend now vs how long monetisation takes (ads, pricing, enterprise budgets). 18:53 — “SaaSpocalypse” ~$300B wiped off software stocks on fear that seat-based SaaS collapses into usage/agent-driven economics. Claude Code “agentic workflows” spook the market: if models do the work, why pay the tool tax? Counterpoint: SaaS doesn’t die—it de-rates (from “growth multiple” to “utility multiple”). Lomax: market likely overcorrecting; enterprise adoption is slow and messy. 26:07 — EU–US tech uncoupling (or… vibes?) France moves to ban civil servants from Zoom/Teams/WebEx → pushes homegrown “Visio” by 2027. Germany (Schleswig-Holstein) + Austrian army shifting off Microsoft to open-source alternatives. Group take: sovereignty goal is real, but government-built software ≠ winning strategy; better to back founders + procurement pathways. 32:44 — Spain vs social media Pedro Sánchez pushes: CEO accountability, misinformation/hate speech enforcement, under-16 social restriction. Smart framing: shifts from “free speech” to public health. Pushback: slippery slope risk → censorship-by-proxy debate. 39:20 — Muskonomics: SpaceX + xAI + “data centres in space” Core claim gets roasted: physics/energy/cooling/payload/latency all feel brutal. Bull case (Lomax): if anyone can brute-force iterate at scale, it’s Musk + launch cadence. Bear case (Mads): narrative may be a financial wrapper to justify merging/funding xAI via SpaceX halo. 51:10 — Anthropic Super Bowl ads + OpenAI shade Anthropic pokes OpenAI over ads in AI (“we’d never”). Take: brand landgrab + positioning move; debate whether the ads were funny or cringe. 53:52 — Europe corner: critical minerals reality check EU auditor warns Critical Raw Materials Act targets likely missed (dependency on China still extreme). Problem isn’t geology—it’s permitting + processing + time (10–20 years to mine/start). US hosts rare earth summit; Europe tries to coordinate while still exporting heavily to the US. 56:53 — Deal of the Week Lomax’s portfolio: Portuguese founder Pedro building LLM-driven clinical trial planning → reduces protocol amendments/costs. Raises $52.5M Series A (one of Iberia’s biggest; top-tier EU Series A scale). Dan: January saw 5 new European unicorns (Aikido shoutout highlighted). Mads: new European growth fund Cambara targeting €30–50M checks; €750M raised toward €1B.