The 2026 product cut tells a different story than the standard launch coverage. Forbes counts ARR; we count job titles. The 25 launches in this index did not just ship features — they created role categories that did not exist 12 months prior. Anthropic's October 2024 Computer Use beta seeded the "AI Agent Engineer" job title. OpenAI's Realtime API for GPT-4o crystallized "Voice AI Engineer" as a defined seat. Cursor's Composer mode rewrote the senior software-engineer job description across Y Combinator's entire winter cohort.
Three findings shape the list.
First, the launch-to-job-title lag has compressed to 90 days. LinkedIn's taxonomy added "AI Agent Engineer" within a quarter of Anthropic's Computer Use release; Sierra, Decagon, and Cognition collectively cleared 1,400+ Computer Use-tagged ecosystem roles in the same window. The October 2024 - January 2025 window is the cleanest example of a single product launch catalyzing a horizontal hiring category. Brendan Foody's Mercor placed-volume tracking shows the speed: short-engagement expert RLHF demand spiked 3x within 60 days of every named launch in our top 10.
Second, the geography of the hiring wave widened, not concentrated. Mati Staniszewski's ElevenLabs (London) plus Eric Steinberger's Magic.dev (London) plus David Ha's Sakana AI (Tokyo) plus Arthur Mensch's Mistral (Paris) account for four of the top 25. The London-Warsaw, London-SF reverse-flow, and Tokyo-as-frontier-research-hub stories are direct downstream effects of the launches we ranked. The job categories are global; the hiring engines are not Bay-Area-only.
Third, enterprise-deployment launches now beat consumer-launch hiring waves on durability. Decagon's Customer Concierge replaced 30%+ of frontline customer-service roles at Eli Lilly, Hertz, and Klarna; Harvey's Workflows release drove "Legal AI Operations Manager" as a real seat at every AmLaw 50 firm; Glean's Workplace Agents put a "Knowledge AI Lead" inside Workday, Reddit, and Pinterest central IT. Consumer launches dominate press cycles; enterprise launches dominate hiring cycles.
Methodology
We longlisted 180 publicly announced AI launches between Q1 2025 and Q1 2026, verified 62 against our coverage universe (1,200 companies + LinkedIn taxonomy + Mercor placement data + Levels.fyi offer disclosures), and selected 25. Each launch was scored on five weighted dimensions: Job-Title Creation (30%), Hiring Wave (25%), Ecosystem Pull (20%), Distribution (15%), Compensation Reset (10%). Stealth-mode releases and internal-only enterprise tools (e.g. JPMorgan's IndexGPT) were excluded — only externally available launches qualified. Job-title creation lag: some role categories take 6+ months to enter ATS taxonomies; the index uses lead indicators where lag applies. Year-over-year anchor: +47% net-new AI role categories on LinkedIn versus 2024 baseline.
The full ranked table is below. Three deep-cut profiles follow.
Anthropic's Computer Use is the single most consequential agent-category launch of the cycle. The October 2024 beta inside Claude 3.5 Sonnet changed the answer to the question "what is an AI agent?" from a research demo to a product. Within 90 days, Sierra, Decagon, Adept-survivors, and Cognition collectively cleared 1,400+ Computer Use-tagged ecosystem roles. Anthropic itself opened 87 net-new roles tied to agent safety, evaluation, and deployment. The April 2026 senior-research reset memo cited agent-safety hiring competition explicitly — the comp band moved because Anthropic, Sierra, and Decagon were bidding against each other for the same 200 humans on the planet who had shipped a deployed agent.
Anysphere's Cursor is the highest revenue-per-employee story in AI startup history. Composer mode, launched in early 2025, is the single product release most cited in 2025 software-engineering job-description rewrites. The team of 60 cleared $200M+ ARR — a number that produced a Series C at $2.5B post-money in Q4 2025. The hiring discipline is the story: Anysphere rejects more senior IC candidates per role than any AI startup outside the frontier labs. The downstream effect is that "AI-first software engineer" or "AI-native developer" is now standard language at GitHub, Vercel, Replit, Stripe, and the entire YC winter cohort. Michael Truell, Sualeh Asif, Aman Sanger, and Arvid Lunnemark built a hiring brand that 50-person companies cannot replicate without their distribution.
Decagon's Customer Concierge produced the most-cited "job displacement" data point of 2025. Jesse Zhang and Ashwin Sreenivas's customer-service agent platform retired or recategorized 30%+ of frontline CS roles at Eli Lilly, Hertz, and Klarna — the data point quoted in three separate 2025 AI policy hearings. The Series C at $1.5B post-money funded an enterprise-sales expansion that reads like the early Salesforce playbook. The other side of the displacement story: Decagon's hiring wave pulled GTM senior leaders from MongoDB, Snowflake, and Datadog, and "Enterprise Agent Architect" is the new senior IC role pulled out of Sierra's wake. The net job-creation balance is positive — but uneven, and concentrated in different geographies than the displaced roles.
The forecast: the Q2 2026 launch cycle will produce two more job categories that do not currently exist. Watch the agent-orchestration launches at Sierra, Decagon, and Anthropic for "Agent Orchestration Engineer" and the long-context-coding launches at Magic, Cursor, and Cognition for "Context Engineer" — both already showing up in early-stage Series A startup job descriptions. By Q4 2026, both will be defined LinkedIn taxonomy categories with median comp bands disclosed in the next ENTRA Salary Index.
For the full company hubs, see Anthropic, OpenAI, and Anysphere. Cross-reference: Top 30 AI Founders to Watch 2026 and Top 20 Highest-Paid AI Roles 2026.