Applied Agents

Built the most widely used LLM application framework.

Recent: Shipped LangChain 1.0 alongside a $125M raise at $1.25B; LangChain and LangGraph now see ~90M monthly downloads and 35% of the Fortune 500 use the stack.

Watch for: Whether LangGraph plus LangSmith becomes the default production agent runtime as frameworks consolidate.

Data framework for LLMs — the plumbing behind RAG everywhere.

Recent: Publicly argued in 2026 that the framework era is over and context quality is the new moat, repositioning LlamaIndex around agentic document workflows.

Watch for: How LlamaIndex survives agent loops absorbing the scaffolding layer it helped create.

Shipping AI-native coding at scale. Replit Agent writes full apps.

Recent: Raised $400M at a $9B valuation in March 2026 — six months after being valued at $3B — and projects $1B revenue by end of 2026.

Watch for: Whether Replit's non-developer audience holds as Cursor, Lovable, and Cognition converge on the same buyers.

Bet HubSpot's product future on AI agents for GTM.

Recent: Runs agent.ai (a marketplace for AI agents with 230K+ registered users) and the simple.ai newsletter (2M+ subscribers) alongside HubSpot's agent push.

Watch for: Whether agent.ai becomes real distribution for third-party agents or stays a hobbyist directory.

Co-founded DeepMind, now runs Microsoft's consumer AI and Copilot.

Recent: Unveiled seven in-house MAI models at Build 2026 and heads the MAI Superintelligence Team formed in November 2025 — Microsoft's hedge against OpenAI dependence.

Watch for: How far MAI models displace OpenAI's inside Copilot, and his claim that most white-collar work automates within 18 months.

Built the fastest-adopted AI coding tool in history from a four-person MIT team.

Recent: Sold Cursor to SpaceX for $60B in August 2026 after hitting $3B ARR in May; stays on as CEO of the subsidiary.

Watch for: Whether Cursor keeps its neutral multi-model stance inside Musk's orbit, or tilts toward xAI models.

Runs Devin, the first credible autonomous software engineer, and absorbed Windsurf.

Recent: Raised $1B at a $26B valuation and says Devin now writes 95% of Cognition's own code.

Watch for: Whether Devin's autonomy pitch survives contact with enterprise codebases better than IDE-shaped rivals.

Turned AI search from demo to daily habit, now pushing into agentic browsing.

Recent: Launched Perplexity Computer in early 2026 — an agent product that orchestrates local models on-device for sensitive files — after a $20B valuation.

Watch for: His 'token value per watt per user' thesis, and whether Comet/Computer survive Google and OpenAI shipping the same thing.

Ex-Salesforce co-CEO building enterprise customer-service agents; also chairs OpenAI's board.

Recent: Raised a $950M Series E in 2026, scaled Sierra to ~700 employees and $200M revenue, declaring 'the era of clicking buttons is over.'

Watch for: Outcome-based agent pricing becoming the enterprise software norm — Sierra is the test case.

Ex-litigation associate who made legal AI the flagship vertical-agent category.

Recent: Passed $300M ARR (from $100M in August 2025) at an $11B valuation, with two-thirds of the AmLaw 100 as customers.

Watch for: Whether Harvey's model-agnostic turn (adding Anthropic alongside OpenAI) widens into a full platform for professional services beyond law.

Ex-Google distinguished engineer who owns the enterprise-search-to-work-agents wedge.

Recent: Said in May 2026 that AI is far from replacing workers and Glean has no plans to cut headcount with it — a contrarian position among agent vendors.

Watch for: Glean's agent platform versus Microsoft Copilot in the fight for the enterprise knowledge layer.

Rebuilt Notion around AI agents; 94% of the Forbes AI 50 run on it.

Recent: Notion 3.0's custom agents pushed AI to 60% of the business (3x YoY) as the company passed $500M ARR with revenue accelerating seven straight quarters.

Watch for: Whether Notion's shared-context 'agents in the workspace' bet beats standalone agent products.

Made voice the interface for AI agents — ElevenLabs powers most production voice agents.

Recent: Passed $500M ARR in May 2026 with roughly 400 employees, driven by conversational agents for customer service.

Watch for: Whether ElevenLabs holds the voice layer as OpenAI and Google bundle native speech into their agent stacks.

Bootstrapped the most profitable image model company in the world — no VC, ~60 people, ~$300M revenue.

Recent: Handed the CEO title to COO Tarun Nandwani on August 1, 2026, keeping himself pointed at research and product.

Watch for: Whether Midjourney's video and world-model work can compete against far better-capitalized labs without raising.

Took AI video from research demo to Hollywood tooling; now chasing general world models.

Recent: Raised toward $860M total at a $5.3B valuation and is expanding Runway's world models into gaming and robotics.

Watch for: His claim that AI lets Hollywood make 50 films for the cost of one $100M blockbuster — and whether studios bite.

CERN physicist turned founder of the fastest-growing software startup in history.

Recent: Lovable was valued at $13.3B in August 2026, roughly doubling from December 2025, after hitting $100M ARR in its first eight months.

Watch for: Whether vibe-coded apps graduate to production software — his 'AI breaks software monopolies' thesis depends on it.

String theorist building Droids — enterprise software agents that run outside the IDE.

Recent: Closed a $150M Series C at $1.5B in April 2026 led by Khosla Ventures, with NVIDIA, Adobe, EY, and Morgan Stanley as customers.

Watch for: His 'dark factory' thesis — software that builds itself with humans only reviewing — against Cognition and Cursor in the enterprise.

Created Claude Code, the tool that made terminal-native agents the default way engineers work.

Recent: Says he hasn't hand-written code in eight months and some days manages tens of thousands of agents; Claude Code drove an 8x jump in code written at Anthropic in 2026.

Watch for: Claude Code's expansion from engineering into general professional work — the wedge for Anthropic's agent ecosystem.

Built the open-source standard for letting agents drive the web — ~80K GitHub stars.

Recent: Grew the YC W25 project into a funded company ($17M raised) that evaluates and powers web control for general-purpose agents.

Watch for: Whether Browser Use stays the neutral web-actuation layer as OpenAI, Anthropic, and Perplexity ship their own computer-use stacks.

The most-read independent chronicler of applied LLMs; his tools and blog set the vocabulary ('prompt injection') the field uses.

Recent: Shipped Datasette Agent in mid-2026 — approval-gated SQL-writing agents — plus LLM releases adding reasoning traces and server-side tools.

Watch for: His running audit of agent security and prompt injection — the reality check vendors route around.