Quick Pulse:
February 2026 flipped the AI script. Labs shipped agentic systems that do work, not just chat. Highlights: $30B raised, 900M users hit, million-token contexts standard. Timeline below tracks it day-by-day.
Week 1: Security & Foundations (Feb 3-5)
Feb 3 – OpenAI Tunes GPT-5.2 for Speed
OpenAI slashed Standard thinking time on GPT-5.2 after user tests favored quick replies over deep dives. Key win: Restored Extended mode (up to 2x prior depth). Users hit 900M weekly actives mid-month; 2025 revenue topped $13.1B.
Feb 3 – Perplexity Goes Enterprise + Comet Free
Perplexity unlocked org tools, APIs, and global Comet browser downloads. Twist: Comet Plus ($5/mo) shares revenue with publishers—AI that pays creators.
Feb 4 – Mistral Drops Voxtral Speech Models
Mistral’s Voxtral Mini/Realtime hit 200ms latency, 13-language translation, phone-scale (4B params). Edge: Open-source Realtime runs offline—perfect for privacy regs.
Feb 5 – Anthropic’s Claude Opus 4.6 Hunts 500+ Zero-Days
Claude autonomously found 500 high-severity bugs in live OSS. Launch: Claude Code Security with patches + free OSS access. Redefines AI defense.
Week 2: Research & Talent (Feb 12-13)
Feb 12 – Google Supercharges Gemini 3 Deep Think
Upgraded for scientists: caught peer-review flaws (Rutgers math paper), optimized crystal growth (Duke). Access: Ultra subs + API early program.
Feb 13 – Anthropic + CodePath for CS Students
Claude integrates into U.S.’s biggest college CS program. Impact: Shapes next-gen devs’ AI habits early.
Week 3: Funding & Power Models (Feb 19-20)
Feb 19 – Anthropic’s $30B Series G at $380B Valuation
Led by GIC/Coatue; up from $4.1B in 2023. Funds frontier infra post-Dario’s AI risk manifesto. Norway’s $2T fund uses Claude for ESG screens.
Feb 19 – Google Ships Gemini 3.1 Pro
MoE architecture, 1M-token context, 65K output. Tiers: Low/Med/High think modes for devs. Crushes agentic SWE benches.
Mid-Feb – Mistral’s OCR 3 & Devstral 2
OCR 3 nails tables/handwriting ($1-2/1K pages). Devstral 2 (123B/24B): 72.2% SWE-bench, 256K context, image multimodal. Vibe CLI automates code. 17B tokens processed Day 1.
Feb 20 – OpenAI Deep Research Glow-Up
Target sites, editable plans, UI tracking. Plus: Lockdown Mode for execs; retired old GPTs. Infra reset: $600B compute by 2030 (down from $1.4T).
Feb 20 – Perplexity’s Utility Blitz
Patents beta (citation-first), privacy-first Email Asst, flight tracking, 10 new sports (ATP, EPL, etc.), Politicians tracker. Deep Research now on Claude Opus 4.6. Snapchat deal: $400M, 1B users.
Week 4: Agents & Scale (Feb 21-26)
Feb 21 – Claude Code Gets Isolation Boost
Worktree isolation, hooks, env vars, auto-updating previews. Fixes race conditions for prod.
Feb 25 – Anthropic Sonnet 4.6: 1M Tokens Live
Top Sonnet for code/agent/design. Extras: Plugin marketplace, Cowork on Pro, PowerPoint/Excel integrations. Auto-memory + /copy in Claude Code.
Feb 25-26 – Perplexity Computer: 19-Model Swarm
Single prompt → parallel agents (Claude/GPT/Gemini/Grok) with filesystems, 400+ tools, persistent memory. Price: $200/mo Max. “Hey Plex” on Galaxy S26. Digital workers arrive.
Feb 26 – Google Nano Banana 2 Goes Global
Gemini Flash image gen: 4K, multi-char consistency, text accuracy, verticals. Hits 141 countries, Search/Lens/Ads/Flow.
Strategic Shifts: The Big Picture
- Agents Rule: Perplexity Computer, Claude Code, ChatGPT Agent, Gemini workflows—multi-model teams finish jobs.
- Context Explosion: 1M tokens standard (Sonnet/Gemini). Handles 1,500 pages/codebases.
- Scale Wars: Anthropic ($380B), OpenAI ($600B compute), Google (weekly ships), Perplexity (OS pivot), Mistral (efficient/open).
- Extra Context: US-China chip tensions + EU sovereignty push frame it all. AI = infrastructure now.
Welp Magazine | February 2026
Compiled from official notes, earnings, benchmarks. Stay pulsed
