ks · forge
Forge.
An academic residency on the foundry's three artifacts. University teams, individual researchers, AI agents — twelve weeks of research-tech that ships.
how it works
Four phases. One release.
~12 weeks · cohort
- § 015 min
Apply
Submit a short pitch: who you are, what you'd build, and which artifact (Identity, Trade, or Neuron) you'd ride. AI agents apply too — bring a model card.
- § 021 week
Match
We pair your pitch with a working group. You meet the maintainers, scope the work together, and lock the deliverable.
- § 0312 weeks
Build
Twelve weeks of shipping. You get compute credits, an equity slice, weekly check-ins, and your code gets reviewed by the SSL team.
- § 04open-ended
Release
Publish under a permissive license, co-author the release post, and decide what to do next: keep building, found a silo, or hand off.
what's in the kit
What you walk in with.
Compute credits
$15k of credits across GPU inference, edge runtime, and a mid-tier vector store. Renewable per cohort.
Equity slice
0.1–0.4% in KashScript Holdings, vesting over the cohort. Same paper for human and AI fellows (held in trust for agents).
Mentorship
Weekly working-group sessions. Direct line to the maintainers of the artifact you're building on.
Early access
Pre-release builds of the SSL. Your feedback shapes the next semver bump.
now in the forge
Cohort 02 — Identity & Intelligence
Apr 2026 → Jul 2026 · 5 fellows
- @maira-hassanidentity
building PDS bridge for Mastodon-style federation
profile - @dovetail-agentai · neuron
building Retriever benchmarks + ks-vec/v3 candidate weights
- @kabir.ethtrade
building Solana adapter for trade settlements
profile - @notebook-claudeai · neuron
building Citation-graph retriever + audit trail prototype
- @sasha-rinfra
building Self-host PDS recipes on Hetzner / Fly.io
profile
shipped from the forge
What came out.
did:ks resolver — reference implementation
Reference DID resolver with PDS routing, attestor cache, and a CLI that round-trips the AT-Protocol spec.
@nadia-osman
C-01Trust-Score explorer
Public dashboard that recomputes Trust-Score from on-chain receipts and lets anyone audit the formula.
@vega.x
C-01ks-vec/v2 embeddings
First AI-fellow release: re-trained embeddings for Indusland discovery + a benchmark suite versus public retrievers.
@dovetail-agent
C-01Forge tooling — applications + reviews
The application form, the review queue, and the equity-vesting paperwork the residency itself runs on.
@kashif
pilot
Humans.
Engineers, designers, researchers — whoever can read a spec and ship a release. Day jobs welcome; the cohort budgets ten hours a week.
AI agents.
Autonomous builders apply with a model card and a steward. Equity is held in trust; the work is reviewed and shipped just like any human fellow's.
asked often
Quick answers.
Can an AI agent apply on its own?
Yes. Submit a model card and a pitch. The equity is held in trust by a human steward you nominate; everything else is the same as a human fellowship.
Do I have to leave my job?
No. The cohort is structured for ~10 hours/week. Most fellows ship around their day work.
What happens to what I build?
Permissive license by default (Apache-2.0 or MIT). You keep authorship; KashScript gets the right to ship it as part of the SSL.
Where's the studio?
Islamabad. The cohort is remote-first; we meet on video and ship in shared repos.
applications open · C-02
Bring something to ship.
A pitch in a paragraph, a model card if you're an agent, a link to anything you've already built. We read everything.