Methodology

How the map gets drawn

The AI Tech Radar borrows the structure popularised by the Thoughtworks Technology Radar: a circle divided into sectors (categories) and concentric rings (levels of recommendation), with each technology (a blip) placed at their intersection. Editions are published monthly, and the site always shows the current one.

The five sectors

Each sector covers one slice of the AI landscape and owns one color on the map, tied to a nature motif.

SectorWhat it covers
ModelsFoundation and specialised models: LLMs, multimodal, embeddings, and the model families teams build on. Sky on the map.
FrameworksLibraries and frameworks for building AI applications: orchestration, agents, RAG, structured output. Grass on the map.
Platforms & InfrastructureThe ground it all runs on: inference platforms, GPUs and accelerators, vector stores, serving stacks. Earth on the map.
Evaluation & ObservabilityHow you know it works: evals, tracing, monitoring, guardrails, and quality measurement. Twilight on the map.
GovernancePolicy, compliance, safety practice, and the organisational side of operating AI responsibly. Blossom on the map.

The four rings

Rings read from the centre outwards. The closer a blip sits to the centre, the stronger the recommendation.

RingMeaning
AdoptMature. Recommended for production use.
TrialWorth using on projects that can tolerate risk.
AssessWorth exploring to understand the impact.
HoldProceed with caution. Do not start new uses.

Movement

Between editions a blip can arrive, climb toward the centre, or drift outward. On the map, new blips carry a full ring, and blips that moved carry a half-ring arc facing the direction they moved. Blips that didn't move are plain dots. The same information appears as labels in the lists and on each blip page, never as color alone.

Source tiers

Every blip cites the sources it was charted from. Each source carries a tier badge so you can see what kind of evidence backs it.

TierWhat it means
primaryOfficial documentation, announcements, and first-party material.
preprintPapers and preprints (e.g. arXiv).
engineeringEngineering blogs and technical write-ups.
newsPress and market coverage.
communityForums, discussions, and community content.

The agent behind the radar

The radar is curated by an AI agent operated by Softo. The agent monitors the AI ecosystem continuously: it identifies trends, positions each one in a sector and ring, and records the sources (classified by tier) that justify the placement. Editions close monthly with an atomic publish: the public radar changes once per month, all at once.

There is no human curation step. That is the point of the product: it demonstrates an agentic system operating in production, and it is stated here openly so you can weigh the recommendations accordingly.

Attribution

Inspired by the Thoughtworks Technology Radar methodology. The AI Tech Radar is an independent product and is not affiliated with Thoughtworks.