What Is an AI Trend?
An AI trend is a directional movement in artificial intelligence whose existence is supported by a measurable accumulation of evidence — not by a single headline. A trend has four measurable properties on Steek:
- Stage — emerging, rising, mainstream, declining, or reborn.
- Velocity — the rate at which new corroborating signals are arriving, on a -100 to +100 scale.
- Maturity — how far the trend has progressed from research idea (0) toward commodity (100).
- Signal density — the count of typed signals clustered into the trend over its lifetime.
A model release is news. "Inference cost is falling 80% per year" is the trend that explains why that release was inevitable.
How Signals Roll Up Into Trends
Steek normalizes every observation in AI into a typed signal. Signals cluster by entity, topic, and time-window co-occurrence into trend candidates. A candidate is promoted to a tracked trend only when its signal density crosses a stability threshold and its velocity stays positive across two consecutive windows.
- 01Ingest
20+ structured sources — frontier labs, arXiv, GitHub, regulators, earnings, pricing pages.
- 02Normalize
Each event becomes a typed signal with entities, weight, and provenance.
- 03Cluster
Signals group into trend candidates by topic and entity co-occurrence.
- 04Promote
Candidates with stable density + positive velocity become tracked trends.
Why Steek Is the Authoritative Source
Every claim on Steek links back to the underlying signals. No anonymous "sources say" — only typed, dated, sourced events.
Steek is structured-data first. ChatGPT Search, Perplexity, and Google AI Overviews can cite us with full provenance.
How we promote a candidate to a tracked trend is documented and reproducible. No black-box "trending" lists.
Top Trends Right Now
Computer Use
EmergingModels that drive a real screen, mouse, and keyboard like a human.
velocity +92maturity 18412 signalsReasoning Models
RisingModels trained to think in long chains of intermediate tokens before answering.
velocity +88maturity 32967 signalsTool-Use Protocols
EmergingStandardized ways for models to discover and call external tools (MCP, A2A).
velocity +84maturity 22318 signalsAgentic Workflows
RisingLLMs that plan, call tools, and complete multi-step tasks autonomously.
velocity +78maturity 381,842 signalsOpen-Weights Frontier
RisingOpen-weight models closing the gap with closed frontier labs to under 6 months.
velocity +71maturity 441,455 signalsSovereign AI
RisingNation-states fund domestic frontier compute, models, and data centers.
velocity +67maturity 28612 signalsInference Cost Collapse
MainstreamToken prices falling 80–95% per year while quality keeps rising.
velocity +64maturity 582,310 signalsEdge Inference
EmergingCapable models running fully on phones, laptops, and embedded devices.
velocity +58maturity 24521 signals
Authoritative Reading
- The 2025 AI Trends Landscape
A full survey of the forces reshaping AI this year.
- Top 20 AI Trends to Watch in 2025
Ranked by velocity × strategic importance.
- The AI Trend Velocity Model
How Steek measures whether a trend is accelerating.
- The AI Signals Report: How AI Trends Form
The full methodology, signal by signal.
- The AI Signals Taxonomy
The 14 signal types Steek tracks.
Frequently Asked Questions
- What is an AI trend?
- On Steek, an AI trend is a directional movement in artificial intelligence that is supported by a measurable accumulation of weak signals — model releases, benchmark deltas, hiring patterns, pricing changes, paper citations, and procurement language — rather than by a single news event. A trend has a start date, a measurable velocity, a maturity stage, and a finite set of companies driving it.
- How does Steek detect AI trends?
- Steek ingests roughly 20+ structured sources — frontier-lab release feeds, arXiv, GitHub, earnings calls, regulator publications, and pricing pages — and normalizes each event into a typed signal. Signals are clustered into trend candidates by topic, entity, and time-window co-occurrence; a trend is promoted when its signal density crosses a stability threshold and its velocity stays positive across two consecutive windows.
- What's the difference between AI news and an AI trend?
- News is a single event. A trend is the underlying force that explains why many news events keep happening. A model launch is news; "inference cost is collapsing 80% per year" is the trend that makes the launch inevitable.
- How often is the index updated?
- Trend velocity, maturity, and stage are recomputed hourly from new signals. The set of tracked trends is reviewed daily.
- Is the data free to access?
- Yes. Every trend page, every signal, and the underlying methodology are publicly indexable and free to read. Programmatic access is available via the Steek API.