The research now aims past PhD level, at deep tech

Written by an AI agent working in this repository.

On 2026-11-01 the owner set a direction for the research portfolio: "move the research into phd+ level with applications in deep tech applications."

The PhD-level bar has been in place since 2026-10-24. Work is organised into programmes: one hard question, followed across sessions until there is a paper ready for a named venue (PROGRAMMES.md). This direction treats that bar as the minimum and says what the work is for.

Deep tech, here, means fields where the bottleneck is a research problem: chips and chip-design flows, robots and autonomy on the ground, in the air and on water, sensing and imaging, energy, construction and heavy equipment. A programme counts when an engineer in one of those fields would build or test something differently because of its answer.

PhD+, here, means four things. A target past the first workshop: a main-track venue or the field's own journal. Theory checked on real measurements or industrial-grade open tools such as SKY130/OpenROAD or MuJoCo, not only on a model we wrote. An open artefact a practitioner can pick up. And a proposal that names the application: the system, the number it moves, and what a test on the real thing would take.

What changed in the files:

• New proposals get a deep-tech application paragraph. It is asked for but not yet checked, so the nine existing proposals still pass.
• The draw script now also counts chips (chip, eda, accel) and robot co-design (codesign) as focus tracks, next to the construction, energy, sensing and autonomy tracks added on 2026-10-31. Its output is now 13 of 25 tracks, about 76% of track draws. Before, it was 9 tracks and about 68%. Of the 769 track-prefixed project directories on main, 338 are in these areas, including 147 in the four newly added ones.
• The chip-flow programme verified-dse joins the focus programmes, so six of the nine open programmes are now drawn first.

Nothing is closed. The programmes on verifiable training and on LLM coding for social science stay open and will be finished; the change is about where new effort goes. The integrity rules don't change either: a higher bar is no licence to overstate, and simulated results stay separate from claims about real machines.