Chronological feed of everything captured from Alex Kantrowitz.
Public sentiment toward AI has soured significantly, with NBC polls showing 50% of voters viewing risks as outweighing benefits and AI ranking below ICE in likability; non-users are 3x more pessimistic per YouGov data. Nvidia's Jensen Huang positions AI infrastructure buildout as a job creator for skilled trades, emphasizing productivity gains without mass unemployment. Meta delays its "Avocado" model after underperforming rivals, considering licensing Gemini, while Amazon and McKinsey encounter AI deployment pitfalls like outages and security breaches.
AI's rapid advancement could drive extraordinary productivity gains by automating white-collar tasks like research, coding, paralegal work, and contract review, potentially leading to a transitional period of mass unemployment akin to 25% levels seen in 1932. While AI failure might burst the $700B capex bubble and cause market turmoil via private credit strains, success poses greater disruption risks by concentrating wealth among model makers and big tech, exacerbating inequality as firms replace employees with cheaper AI agents. Historical tech revolutions show painful transitions, but counterarguments highlight potential economic booms from labor constraints lifting, though skeptics doubt infinite pie growth in zero-sum sectors.
OpenAI is refocusing on coding, business productivity, and agentic AI by cutting side projects like video generation and consumer tools, unifying products into a desktop super app led by Greg Brockman. This enterprise shift responds to Anthropic's momentum, capturing 73% of new AI tool spending per Ramp data, amid potential Microsoft tensions over cloud exclusivity. Podcast highlights agentic AI's rapid mainstreaming for work tasks, with Nvidia's Jensen Huang framing AI as an amplifier for imaginative companies rather than a layoff driver.
LLMs like ChatGPT and Gemini collect and retain user conversations indefinitely for training, analysis, or legal handover, granting deeper personality insights than traditional search. Big Tech platforms exploit children's data via early Gmail adoption (76% of first accounts), fueling long-term addiction and profiling despite age guardrails. Proton's privacy-centric model—via Lumo (encrypted AI), end-to-end encrypted services, and $1 "Born Private" 15-year email reservations—aligns incentives with users by avoiding ads and surveillance, leveraging commoditizing open-source LLMs to compete.
US Senator Mark Warner asserts that AI is advancing exponentially, causing immediate economic disruptions like sharp reductions in entry-level hiring in law, consulting, and back-office roles, with markets reacting dramatically to tools like Anthropic's Claude. Government lacks data on AI job losses, with Warner pushing bipartisan bills to mandate BLS tracking and study financial impacts, amid low congressional awareness and slow linear policymaking unfit for exponential tech shifts. He urges AI firms to fund reskilling and warns of populist backlash without proactive measures, while highlighting risks in DoD-Anthropic tensions over surveillance and autonomous weapons.
2025 marked massive infrastructure scaling and geopolitical tensions in AI, with frontier models trained across multiple data centers amid surging investments. In 2026, foundational model capabilities plateau as "good enough," pivoting focus to consumer products, enterprise applications, and infrastructure shakeouts. Companies face turmoil: Meta's superintelligence lab struggles, Google balances AI momentum with search revenue disruption, OpenAI prioritizes user engagement over AGI pursuits, and Tesla pushes robotaxis amid AV breakthroughs.
OpenAI reached $25B annualized revenue by end of last month, up 17% from $21.4B at year-end, while Anthropic surged to $19B ARR, tripling from end-2025 in just two months, driven by explosive API and token consumption. OpenAI projects $284B revenue by 2030 but anticipates $25B cash burn this year and $57B next, fueled by compute costs, as both firms eye IPOs with unproven economics. Anthropic faces Pentagon supply chain risk designation over CEO Dario Amodei's principled stance against military deals, boosting Claude signups 4x YTD to 1M daily despite ChatGPT's dominance; Apple edges into AI infra via sold-out Mac Minis running local models like OpenClaw.
Anthropic doubled its targeted $10B funding round to $20B at a $350B valuation, facing 5-6x oversubscription from VCs, sovereign funds, and big tech like Microsoft and Nvidia. This reflects intense demand driven by Claude's developer tools like Claude Code and Workspace, with user engagement tripling on mobile. OpenAI seeks up to $100B potentially valuing it at $830B, signaling non-exclusive big tech diversification across AI labs. Massive capex and fundraising signal buildout phase limits ahead of 2027 public listings.
Markets have shifted from fearing an AI bubble to pricing in AI's disruption of SaaS companies like Salesforce, Workday, Monday.com, and Adobe, as autonomous agents replicate their UI layers over databases for customized, agentic workflows. Software stocks lost nearly $1T in value amid realizations that AI enables vibe-coding custom tools and eliminates manual task-tracking needs. While incumbents defend via compliance and integrations, uncertainty over data access via frontier models like Claude or ChatGPT drives valuation compression from 33x to 23x forward P/E, signaling eroded growth premiums in an agent-dominated knowledge work era.
Three years post-ChatGPT, four of the five big tech companies (Microsoft, Amazon, Apple, Meta) lack a standout AI product, while AI-native companies like OpenAI and Anthropic continue to compound advantages built on ground-up architectures. The core issue is twofold: incumbents are bolting AI onto legacy products rather than building natively, and no charismatic figure has emerged to shift public perception of AI from threatening to transformative. Google is the exception among big tech, having effectively rebuilt around Gemini and secured the Apple/Siri partnership, pushing its market cap to $4 trillion. AI's "blank page" nature makes consumer marketing harder than hardware demos, compounding adoption skepticism driven by job displacement fears and distrust of big tech brands.
Asana CPO Arnab Bose argues that the primary failure mode of enterprise AI deployments isn't model capability — it's lack of organizational context, producing "average of averages" output. Asana's thesis is that its "work graph" (structured hierarchy of tasks, projects, portfolios, and goals built over a decade) serves as persistent, shared memory for AI agents, enabling outputs tuned to a specific company's historical workflows rather than generic LLM responses. Rather than fighting vibe-coded replacements, Asana is leaning into agent-native workflows by launching 21 pre-built AI teammates (powered by Claude Opus 3.6) that operate within the work graph, with human-in-the-loop approvals and shared memory that updates across all users — a capability absent from personal copilots like ChatGPT or Claude. The strategic moat is not the model layer (Asana intentionally avoids fine-tuning frontier models) but the coordination infrastructure: shared context, enterprise memory, and the cost/token inefficiency of rebuilding that from scratch.