Divya van Mahajan

The Month AI Woke Up

ai-daily-brief-podcast

The Month AI Woke Up — Study Document

Overview

This episode of the AI Daily Brief (published March 2, 2026) serves as a monthly recap of February 2026, framed as the month that AI meaningfully “woke up” — a period in which autonomous AI agents became practically useful, Wall Street began pricing in AI disruption to white-collar industries, and the relationship between Washington and Silicon Valley over AI governance broke into open conflict. The host (name not stated) collaborates with KPMG on these monthly recaps. The episode opens with breaking news on the Anthropic/Pentagon/U.S. Government dispute before transitioning to the broader thematic summary of February.

Source video URL: not provided.

Prerequisites

  • Familiarity with major AI labs: Anthropic, OpenAI, Google DeepMind, ByteDance
  • Basic understanding of large language models (LLMs) and AI agents
  • Awareness of the “vibe coding” / AI-assisted software development trend
  • General knowledge of U.S. government procurement and national security frameworks
  • Familiarity with prior AI benchmarks and the concept of AGI as an industry term
  • Basic understanding of startup fundraising rounds and SaaS business models

Main Points

1. Anthropic vs. the Pentagon — A Contradictory Situation

  • The U.S. government designated Anthropic a “supply chain risk” on a Friday afternoon, theoretically requiring other government contractors to drop Anthropic relationships.
  • Despite this designation, Claude was reportedly used during U.S./Israel strikes on Iran the following Saturday — to analyze intelligence, select targets, and run battlefield simulations.
  • Autonomous Lucas kamikaze drones were deployed in the operation, confirming autonomous weapons are already part of modern warfare, independent of frontier LLMs.
  • OpenAI’s models were not used in the operation; they had not yet been approved for classified settings.
  • President Trump confirmed a six-month phase-out period during which Anthropic’s technology would remain in military use.
  • Democrat Congressman Seth Moulton highlighted the contradiction: either the Pentagon used a declared security risk in combat, or the original designation was not made in good faith.

2. Consumer Sentiment Surge for Anthropic / Claude

  • Amid the controversy, Claude rose to #1 on app download charts, overtaking ChatGPT for the first time.
  • Claude had been outside the top 100 free apps at end of January and mostly outside the top 20 through February.
  • Anthropic promoted a ChatGPT memory migration feature — essentially a large prompt users copy from ChatGPT and paste into Claude to transfer conversational context.
  • Commentary noted this reveals that “memory as a moat” may be weaker than assumed, since memory can be exported as a portable file.

3. OpenAI’s Pentagon Contract — Language and Safeguard Debates

  • Sam Altman hosted an AMA on X, arguing the supply chain risk designation sets a bad industry-wide precedent.
  • OpenAI published its contract with the Department of Defense, including AI red lines:
    • Autonomous weapons: AI will not independently direct autonomous weapons “where law, regulation, or department policy requires human control.”
    • Domestic surveillance: AI shall not be used for “unconstrained monitoring of U.S. persons’ private information” consistent with applicable authorities.
  • Critics (e.g., Peter Wildeford) noted this language effectively permits any lawful use as determined by the Pentagon — meaning if the Pentagon deems a use lawful, OpenAI’s safety stack would be in conflict with the contract.
  • OpenAI’s NatSec lead Katrina Mulligan countered that the same logic applies to Anthropic’s red-line contract language — governments can choose to interpret any contract language permissively.

4. OpenAI’s $110 Billion Fundraising Round

  • OpenAI finalized the largest startup fundraising round in history at $110 billion, valuing the company at $840 billion post-money — the most valuable startup ever and the 15th most valuable company globally.
  • Three corporate strategic investors:
    • Amazon: $50 billion (split: $15B end of March, $35B contingent on IPO or milestones reportedly including AGI)
    • NVIDIA: $30 billion (includes additional chip supply agreement)
    • SoftBank: $30 billion
  • The round remains open; ~$10 billion more expected from UAE fund MGX by end of March.
  • Amazon expanded its AWS server deal with OpenAI from $38B over 7 years to $138B over 8 years, with commitments to use Amazon’s Trainium 3 and Trainium 4 chips.
  • Amazon will jointly develop AI models with OpenAI for Amazon consumer apps; Amazon becomes exclusive provider of OpenAI’s Frontier AI agent management tool (except first-party deployment, which remains on Azure).
  • Microsoft did not invest further but retains revenue-sharing rights, including a cut of OpenAI revenue generated through AWS.

5. ChatGPT User Growth Metrics

  • ChatGPT reached 900 million weekly active users (up from 800 million in October 2025).
  • 50 million subscribers — January and February described as the largest months of new subscriber growth in history.
  • 9 million+ paying business users across startups, enterprises, and governments.
  • Weekly Codex users tripled since the start of the year to 1.6 million.

6. The AI Insider Awakening — Coding Agents and Autonomy

  • The period from late November 2025 through February 2026 represented a qualitative shift in AI coding agents: they went from largely non-functional for autonomous tasks to reliably completing large, long-horizon tasks.
  • Former OpenAI co-founder Andrej Karpathy articulated this publicly: coding agents “basically didn’t work before December, and they basically do now,” with models showing higher quality, long-term coherence, and tenacity.
  • Karpathy described the new paradigm: instead of typing code into an editor, developers now spin up AI agents, give them a goal in natural language, and manage their work.
  • The key metric in this era is orchestration: how many agents can be run simultaneously to produce meaningful, compounding output.

7. OpenClaw — The Face of the Agentic Era

  • OpenClaw (originally named ClaudBot, briefly MoltBot) emerged as the clearest product manifestation of increased autonomy ambition.
  • Built externally, OpenClaw gave powerful models access to user systems to perform meaningful autonomous or semi-autonomous work — from inbox management and calendar scheduling to running multi-agent teams.
  • The host built a 10-agent team (1 developer, 2 researchers, 5 project managers, 1 chief of staff) using OpenClaw.
  • Claw Camp, a self-directed learning program the host created around OpenClaw, attracted nearly 5,500 participants despite being technically demanding.
  • Mac Minis and Mac Studios became associated hardware symbols of this agentic era.
  • Anthropic originally asked for a name change from “ClaudBot” rather than embracing the tool — seen as a missed opportunity. OpenAI later hired OpenClaw’s creator to build similar systems internally.

8. The “Clawification” of AI — Industry-Wide Agentic Expansion

  • OpenAI released the Codex app as a direct competitor to Claude Code.
  • Anthropic released Remote Control for Claude Code (manage sessions from mobile) and Scheduled Tasks in Claude Cowork.
  • Perplexity announced Perplexity Computer — same agentic paradigm.
  • Microsoft announced Copilot Tasks; CEO Satya Nadella reportedly using OpenClaw personally and encouraging his team.
  • Notion released custom agents.
  • Non-technical users also adopted these tools: CNBC journalist Deirdre Bosa built her own Monday.com clone in one hour using Claude Cowork.
  • Solo developer Ben Serra built Pulsia — an AI platform for running autonomous AI-powered businesses, reaching $1.25M+ annual run rate within weeks. Users give Pulsia an idea; it builds and operates an online business with access to GitHub, Meta Ads, and other services.

9. Wall Street Wakes Up — The SaaSpocalypse

  • February became the month Wall Street began pricing in AI disruption to white-collar SaaS and knowledge-work industries.
  • The trend started in late January when Google’s Genie 3 demo (60-second immersive world creation) caused gaming publisher stocks to fall.
  • Pattern: each time Anthropic announced a new plugin for Claude Code or Cowork, stocks in the adjacent industry category fell sharply.
    • IBM saw its worst single-day stock drop in 25 years after Anthropic published a blog post about its COBOL tool.
    • Affected sectors: gaming, legal, finance, productivity software.
  • Bloomberg and WSJ both described the “hot new trade” as dumping stocks in AI’s crosshairs.
  • Citrini Research published a viral piece, “The 2028 Global Intelligence Crisis”, outlining a theoretical economic doom loop from AI-driven white-collar displacement.
  • Block announced layoffs of 4,000 employees (~40% of staff), cited by many as early evidence of predicted AI-driven white-collar carnage.

10. Washington Wakes Up — The Broader AI Governance Power Struggle

  • The Pentagon/Anthropic conflict was framed not just as a contract dispute, but as the first major public manifestation of an inevitable power struggle between government and AI labs over who controls how AI is deployed.
  • Core disagreement: Anthropic sought explicit red-line carve-outs (no autonomous weapons, no domestic mass surveillance); the White House pushed for “any lawful use” as the standard.
  • The conflict escalated publicly, drawing bipartisan criticism (e.g., Senator Tom Tillis called it “sophomoric”).
  • President Trump and Defense Secretary Pete Hegseth announced Anthropic would be designated a supply chain risk; a six-month phase-out was confirmed.
  • The episode notes this is “just the first in what will be a much bigger power struggle in the years to come.”

11. Model Releases and Benchmarks in February

  • ByteDance SeedDance 2.0: Video generation model; some in the industry asked whether this was the first Chinese open-weight model to exceed U.S. capabilities in its domain.
  • Anthropic: Released Sonnet 4.6, completing the full 4.6 model suite alongside Opus 4.6.
  • Google: Released Gemini 3.1 Pro (emphasizing multimodal capabilities) and Nano Banana 2 (faster, cheaper, improved text handling and reasoning).
  • Long Horizon Task benchmarks: Both Codex 5.3 and Opus 4.6 scored extremely high, with Opus 4.6 “basically off the charts” — indicating that existing benchmarks can no longer keep pace with model progress.
  • Anticipated in March: DeepSeek 4 release (long-anticipated, repeatedly delayed) and reportedly GPT-5.4 with significant capability improvements.

Key Concepts

  • Agentic AI / AI Agents: AI systems that are given a goal in natural language and autonomously plan and execute multi-step tasks, rather than responding to individual prompts.
  • OpenClaw: An externally built tool enabling users to give frontier AI models access to their computer systems to perform extended autonomous or semi-autonomous work; became the defining product of the agentic AI era in early 2026.
  • Claude Code / Claude Cowork: Anthropic’s developer-focused and productivity-focused AI tools, respectively, supporting agentic and semi-autonomous workflows.
  • Orchestration: The practice of running multiple AI agents simultaneously and coordinating their outputs toward a larger goal; described as the key value-creation challenge in the agentic era.
  • Vibe Coding: Colloquial term for AI-assisted software development where the user describes desired outcomes in natural language rather than writing code directly.
  • SaaSpocalypse: Informal term for the market phenomenon in February 2026 where SaaS and white-collar software company stocks fell sharply as AI’s capability to replace their functions became apparent.
  • Supply Chain Risk Designation: A U.S. government classification that, if applied to a company, could require other government contractors to sever relationships with that company.
  • Red Lines (AI contracts): Explicit prohibitions in AI deployment contracts specifying categories of use that are never permitted (e.g., autonomous weapons, mass surveillance).
  • Long Horizon Task Benchmark: A benchmark measuring AI models’ ability to complete complex, multi-step tasks over extended time periods; emerged as a key progress metric in 2025.
  • SeedDance 2.0: ByteDance’s video generation model, notable for prompting discussion about Chinese open-weight AI models matching or exceeding U.S. capabilities.
  • Citrini Research / 2028 Global Intelligence Crisis: A viral research report articulating a theoretical economic catastrophe scenario driven by rapid AI-induced white-collar displacement.
  • Pulsia: An AI platform built by solo developer Ben Serra that autonomously creates and operates online businesses based on a user-supplied idea.
  • Trainium: Amazon’s proprietary AI training chip, with Trainium 3 and forthcoming Trainium 4 now committed to as part of the OpenAI-Amazon partnership.

Summary

February 2026 was a watershed month in AI, characterized by a cascade of awakenings across multiple groups simultaneously. For the most engaged AI users, the realization crystallized that the generation of models arriving in late 2025 had crossed a meaningful threshold in autonomous capability — particularly in software development, where AI coding agents went from largely unreliable to genuinely functional for large, long-horizon tasks. This shift was embodied by OpenClaw, which gave users a concrete interface to deploy autonomous AI agents and sparked rapid industry-wide imitation. On Wall Street, the same realization translated into market disruption, as investors began aggressively repricing SaaS and knowledge-work stocks in anticipation of AI-driven displacement, culminating in viral research framing the trend as a potential economic crisis. In Washington, the month produced the first major open confrontation between the AI industry and government over who controls AI deployment standards, surfacing in the Anthropic/Pentagon conflict over red-line contract language — a dispute made more surreal by Claude being used in active military strikes hours after being designated a national security risk. Taken together, February 2026 marked the point at which AI’s transformation from a promising technology into a disruptive force reshaping software development, financial markets, and geopolitical power dynamics became undeniable across a broad and previously skeptical public.