Five reports, in two categories. AI covers the software stack, the labs setting its direction, and the frontier bets nobody has solved yet. Robotics covers physical AI, where every deployment claim is graded against a stated evidence bar and mapped onto the supply chain underneath it. Same method throughout: volatility-tiered fact-checks, conflicts logged rather than smoothed, and every figure sourced.
Software AI, from the silicon underneath it to the people deciding where it goes next. Where the value sits, what the builders believe, and who is attacking the problems nobody has solved.
12 layers from silicon to verticals, three cross-cutting spines (security, regulation, energy), and the demand-side record: 88% of enterprises have adopted AI, about 6% can show EBIT impact. Charts, competitive tables, a searchable index and a knowledge graph.
What 22 frontier labs are actually building — where they contradict one another, where they quietly agree, and who is betting against LLMs entirely.
38 frontier bets sorted by unsolved problem — Skyfall, Mirendil, Oak Lab, Ineffable, Discovery Loop, Ndea. Plus a sortable ledger of 2026's funding, what came out of YC, and the method for tracking it yourself.
Physical AI, where a claim has to survive contact with hardware. What has actually shipped and what has only raised, and the fourteen-layer supply chain neither side can escape.
One robot industry earns billions and is almost entirely teleoperated. The other has raised more capital in eight months than the first earns in a year, and mostly cannot say how many units it has in the field. Every deployment claim graded against a stated evidence bar — plus the fortnight the US closed its market to Chinese robots and Unitree went public.
14 layers from rare-earth magnets to robot-delivered services, mapping the entire industry — humanoids, autonomous vehicles, drones and defence, surgical, agricultural, warehouse and industrial. Competitive tables, a searchable company index, a force-directed knowledge graph, and strategic build opportunities by funding tier.
The other half of the shelf: things built rather than researched. Working software and the thesis behind it, each with its own source.
An AI advisor for personal finances — the registered-account rules, the contribution room, and the sequencing decisions, handled conversationally instead of in a spreadsheet.
The landing page for the product, and the market thesis underneath it — who the buyers are, where the wedge sits, and why consent is the thing worth owning.
you're deciding where to build or invest. It ranks every layer's players, sizes the markets, and flags the gaps worth funding, from bootstrappable to deep-pocketed.
you're deciding what to believe. It maps the strategic bets of the 22 labs that set the direction the entire stack has to follow.
you're deciding what to watch. It sorts the newest labs by the problem they claim to be solving, follows the money month by month, and lists the dated events that would prove any of it.
you need the map. 14 layers and 181 companies from rare earths to robotaxi services, with a searchable index and a knowledge graph — the reference document for who sits where in physical AI.
you want to know what is actually running. It grades every physical-AI deployment claim against a stated evidence bar, separates the robot industry that earns from the one that raises, and maps the supply chain neither can escape.