A comprehensive, 205-slide analysis of where agentic AI stands today—the money, the technology, the winners, and what comes next. By Jon Radoff.
What's Inside
- The Money: $211 billion in AI venture capital—50% of all global VC—and the mega-deals reshaping the industry
- The Technology: 1,000× cost reduction, task horizons doubling every 123 days, and open-source models closing the gap to within 1%
- The Evolution: From vibe coding to agentic engineering—how coding agents went from novelty to 4% of all GitHub commits
- The Creator Economy: 100,000+ new products built daily on AI-native platforms, and the rise of the “1000x developer”
- Who Is Winning: The SaaSpocalypse, $2 trillion in evaporated SaaS market cap, and why API-first composable products are taking over
- Infrastructure Wars: MCP protocol adoption at 97M+ downloads/month, agent-friendly design as the new competitive moat
- Direct from Imagination: AI 3D generation, world models, and the convergence of games, spatial computing, and agentic AI
- Beyond Engineering: How AI agents are transforming legal, healthcare, finance, education, and scientific research
- Robotics & Embodied Agents: The humanoid robot landscape and the disappearing line between software and physical agents
- Machine Societies: Multi-agent cooperation, autonomous agent risks, and the governance gap
Read the Deck
Flip through the full presentation below, or open it in a new window.
Licensed under Creative Commons Attribution 4.0
Why This Report Matters
Agentic AI is no longer a research curiosity. In 2025, AI venture capital hit $211 billion—capturing half of all global venture funding—and the agentic AI market alone is projected to reach $93 billion by 2030 at a 65.5% compound annual growth rate. This presentation maps the entire landscape: from the unprecedented capital flows into frontier models and coding agents, through the technology breakthroughs making autonomous AI practical, to the real-world business impact already visible across every industry.
The numbers tell a story of acceleration that has no historical precedent. The cost of running frontier AI models has dropped 92% since GPT-4's launch in March 2023. Agent task completion horizons—how long an AI can work autonomously on a complex task—are doubling every 123 days. Claude Opus 4.6 can now sustain 14.5 hours of autonomous work. Open-source models have closed the performance gap with proprietary systems to within a single percentage point, down from a 17% deficit just two years ago. These converging trends mean that the capabilities available to every developer and creator are expanding at a rate that makes traditional product planning cycles obsolete.
The Rise of Agentic Engineering
One of the most significant shifts documented in this report is the evolution from “vibe coding” to what is now called agentic engineering. In early 2025, developers discovered they could describe software in natural language and watch it materialize. That initial burst of excitement was real, but it hit walls at production scale: technical debt accumulated, security gaps appeared, and unmaintainable code proliferated. By 2026, the practice matured. Agentic engineering keeps the creative energy of conversational development but adds the oversight, architecture, and production discipline required for real systems. The human role has shifted from writing code to orchestrating AI agents—a change that Cursor's $1 billion ARR in 24 months and Claude Code's contribution to 4% of all GitHub commits make undeniably concrete.
The SaaSpocalypse and the New Winners
Between January 15 and February 14 of 2026, $2 trillion in SaaS market capitalization evaporated. The root cause is structural: when AI agents can perform the work of dozens of human users, per-seat licensing collapses. Ten agents replacing a hundred sales representatives means 90% fewer Salesforce seats. This is not a market correction—it is a permanent repricing of how software creates and captures value. The winners that emerge from this upheaval share a common trait: they are API-first, CLI-composable, and agent-friendly. Companies like GitHub, Stripe, Supabase, Vercel, and Cloudflare are thriving precisely because agents can use their products programmatically. Meanwhile, UI-heavy, seat-based platforms that depend on humans clicking through workflows are seeing their business models structurally undermined.
The Creator Economy Explosion
AI-native creator platforms have produced some of the fastest revenue growth in software history. Cursor went from zero to $1 billion in annual recurring revenue in 24 months—the fastest B2B SaaS ramp ever recorded. Lovable reached $200 million in eight months. Bolt.new hit $40 million in four weeks. These platforms are enabling a new category of builder: people who have ideas and domain expertise but not necessarily years of programming experience. Natural language has become a compiler, and the barrier to building software has shifted from technical skill to imagination and intent. The report documents how 100,000+ new products are being built daily on these platforms, with 41% of all code now AI-generated.
Infrastructure, Protocols, and the Agent-Friendly Moat
The Model Context Protocol (MCP) has emerged as a critical piece of AI infrastructure, with 97 million SDK downloads per month and adoption by every major AI lab. MCP solves the integration problem—instead of requiring custom connections between every model and every tool, it provides a universal standard. Gartner projects that 40% of enterprise applications will have MCP-enabled agents by the end of 2026. The report argues that agent-friendly design is the new competitive moat: if agents cannot use your product via API, they will use a competitor's. This principle is already reshaping how infrastructure companies, developer tools, and even databases compete for market share.
Beyond Software: Industry Transformation
The final sections of the report extend the analysis beyond software engineering into every major industry. In legal, AI now performs document review 80 times faster than humans. In healthcare, over 200 AI-designed drugs are in clinical development, with first approvals expected in 2026-2027. Goldman Sachs reports 3-4x productivity gains and can generate IPO prospectuses in minutes instead of weeks. In education, Khan Academy's Khanmigo has grown from 40,000 to 1.4 million users in two years. The robotics section documents the humanoid robot landscape—from Figure AI's $39 billion valuation to Unitree's $13,500 consumer humanoid—and the convergence of software agent architectures with physical embodiment.
Whether you are a founder evaluating where to build, an investor assessing where capital should flow, or a technologist trying to understand what the next twelve months will bring, this report provides the data and analysis to inform those decisions.