Four vectors.
One field of change.
A vector, in the Index, is a direction of structural change — a trajectory that independent technical developments keep aligning along. Every dossier we publish is plotted on exactly one vector, so the Index reads as a system: follow a vector and you follow one axis of where the field is going.
Four vectors cover the territory we index. Each one below explains what it covers, the signals that put a topic on our research queue, and the dossiers indexed on it so far.
The physical substrate of intelligence.
Compute tracks the hardware layer: silicon, advanced packaging, accelerators, quantum machines, and — increasingly — the power and thermal systems that decide what can actually be built. As AI demand collides with grid capacity and process-node economics, the constraints on computing are becoming physical again.
Dossiers on this vector follow the money and the megawatts: who is fabricating what, how chips are assembled and interconnected, and where the energy to run them comes from.
Signals we track
- Process-node and advanced-packaging roadmaps (chiplets, UCIe, 3D stacking)
- Accelerator and neuromorphic architectures beyond the GPU
- Data-center power: SMRs, grid interconnect queues, on-site generation
- Thermal limits — cooling, density, and the end of free scaling
Dossiers on this vector
The architecture connecting the silicon.
Systems covers core software and architecture engineering: kernels, interconnects, memory standards, networks, and the distributed designs that turn racks of hardware into one coherent machine. This is where thirty-year-old assumptions — like the memory wall, or the private memory bus — get quietly dismantled by standards work.
Dossiers on this vector read standards filings and architecture papers so you don't have to, and trace which of them will actually restructure how systems are built.
Signals we track
- Interconnect and memory standards (CXL, UALink, disaggregated memory)
- Open ISAs and their certification paths (RISC-V in safety-critical systems)
- Kernel, scheduler, and network-fabric evolution at datacenter scale
- Storage and memory hierarchies under AI working-set pressure
Dossiers on this vector
The trust layer under everything else.
Security tracks cryptography, zero-trust protocols, supply-chain integrity, and the protocol-level vectors attackers actually use. The structural story here is a migration: from perimeter trust to verified identity, from classical ciphers to post-quantum ones, from "trust the vendor" to "verify the bill of materials."
This vector is queued: the taxonomy is live, topics are in research, and the first dossier publishes once it clears the Argus floor like everything else.
Signals we track
- Post-quantum cryptography migration timelines and standards (NIST PQC)
- Hardware roots of trust and confidential computing
- Software supply-chain integrity — SBOMs, signing, provenance
- Zero-trust architecture rollouts beyond the buzzword
Dossiers on this vector
The languages writing the next layer.
Syntax follows LLMs, compilers, interpreters, and the automated codebases now writing meaningful fractions of the world's software. The structural question isn't whether machines write code — they do — but how languages, toolchains, and verification practices reshape themselves around that fact.
This vector is queued: topics are in research, and the first dossier publishes once it clears the Argus floor like everything else.
Signals we track
- Model-architecture shifts that change inference economics
- Compiler and toolchain integration of learned components
- Agentic code generation and the verification gap it opens
- Language evolution under machine authorship
Dossiers on this vector
Read the dossiers plotted on these vectors.
Enter the Index →