Roadmap
Here is where Noolog actually stands — not a pitch deck. Now is pinned at the top. Fold the timeline into the past to see what shipped, or into the future to see what's coming. Same descent, you just choose the direction of time.
We would rather under-promise and out-build. Everything past "now" is a sequence gated on adoption, not a calendar — each phase unlocks when the one before it earns it. Nothing marked planned exists yet, and we say so on every line. That is how we actually plan.
Now · in progress
The fusion model is patch theory — commutable, addressable hunks; adopt = cherry-pick; provenance = a patch DAG. We are making that literal: a git-3way composition engine so every cycle is real, reviewable commits, with Pijul evaluated as a native patch-algebra backend for when drops turn dependent.
Unlocks: Code development
The fusion's edits become a first-class, reversible, content-addressed record on a git substrate.
A second composition engine behind the same trait — chosen on real dependent-drop frequency, not hype. Git stays the interop/export layer regardless.
The gated "commit to a fusion" action is the last piece before the app is fully round-trip.
2026 to date · shipped
Paper into product. The NSED orchestrator hit a GA release, and a native desktop app drives it. Real, in hand, today.
Unlocks now: Use AI like a CEO · Zoom for agents
Fusion engine, OpenAI-compatible /v1/responses API, terminal UI, and CLI (nsed serve / status / trace / run / init). Now on the quorum-rs v0.7 line, nearing v0.8.
Per-operator credit / debit / earning ledger with 402 budget enforcement — the metering a marketplace needs.
Commitment chain with agent signing, Poseidon hashing, and nsed-crypto-core primitives (Ed25519, ML-DSA-65). Every answer is a chained record, not a black box.
Slint, Rust-native. Assembly-picker onboarding, live fusion feed, evaluation-rubric radar, convergence sparkline, transcript export, and the send path into a fusion. Read-side is live; the write path is in progress.
The demand test you are reading, with a pre-order flow across the hardware tiers.
January 2026 · shipped
Peer-reviewed and out in the open: NSED — many minds reasoning as one. On the AIME 2025 test an ensemble on consumer hardware scores 84% — matching a top single model and beating a simple vote of the same models by +30 points. This is the drum every layer marches to.
Published January 2026 — NSED, arXiv 2601.16863. The foundation everything else is built on.
2025 · origins
Before Noolog there was a question: can a crowd of peers rank merit and share ownership with no central referee? Peeramid Labs' 2025 experiments on decentralised intellectual property and meritocratic, game-theoretic ranking answered it — the root system the fusion engine still runs on.
The case for peer-ranked, game-theoretic intellectual property — Peeramid Labs research, 2025.
On-chain meritocratic ranking, in the open — Peeramid Labs, 2025.
These are directions gated on adoption, not a calendar. The order is real; the timing is not promised. Each phase opens only when the one before it earns it.
Next · the next push
The near horizon: a JMAP mail / drive / agent client as the way in, clients on every surface, and the git + libpijul change-tracking system carrying live fusion history. The JMAP transport foundation already runs internally; the secured, customer-facing product is not built yet.
A familiar mail-and-drive surface as the onboarding wedge, over JMAP-on-:443 encrypted mailboxes (a Stalwart client path).
Planned — not yet shipped; early transport groundwork exists.The same Slint (Rust) client on iOS and Android — winit + femtovg, one codebase across desktop, mobile and browser.
Planned — not yet built.Today's in-progress work matured into the shipping record, with libpijul as the second composition engine behind the same trait.
Planned — a direction, not a scheduled release.Self-hosted, multi-organisation forge so teams own the repos their fusions live in.
Planned — not yet stood up.Extend the paper's result past AIME — mixture-of-models scored across many domains.
Planned — a research direction, not a scheduled release.After that
Prove who you are without handing over who you are, and let nodes find and trust each other with no central directory — zero-knowledge identity plus peer-to-peer federation. All directional; none of it is shipped IP yet.
Prove age, nationality, or personhood from a passport or an eIDAS 2.0 wallet, revealing nothing else.
Planned — not yet built.Zero-knowledge proofs over signed email (ZK-DKIM), backed by a Groth16 proving stack.
Planned — a design direction, not yet built.Align the identity layer with the ISO/IEC 27565:2026 zero-knowledge-proof privacy guidance.
Planned — a compliance target, not yet certified.Node-to-node federation over Logos Delivery (formerly Waku) — libp2p transport with RLN rate-limiting, no broker in the middle.
Planned — not yet built.A JMAP mailing-list app model, with peers discovered by PageRank over attestations rather than a directory.
Planned — a direction, not a scheduled release.Then
Harden the wire and settle value on-chain: we already sign with ML-DSA-65 over classical AES-256-GCM encryption; this builds the agility to hot-swap the network to PQ encryption, and puts ranking and agent payments on-chain. Directional, not shipped.
Hot-swap the network to ML-KEM / Kyber encryption and ML-DSA / Dilithium signing without hardware replacement.
Planned — a direction; today's encryption is classical AES-256-GCM.The meritocratic ranking contracts, next generation, on zkSync Era.
Planned — not yet deployed.Machine-to-machine settlement over x402, so agents can pay for compute directly.
Planned — a direction, not a scheduled release.After that
Applications that run against a TEE-secured protocol — the software half of the hardware root of trust, so your context stays sealed on the box that computes it. A design goal today, not shipped IP.
Attested, sealed execution for the apps that run on a node.
Planned — the trusted-execution binding is a design direction, not yet built.Each account's compute isolated in its own sealed enclave — no cross-tenant leakage.
Planned — a design direction, not yet built.Root of trust spread across manufacturers (near.ai / Phala first), so no single vendor is trusted alone.
Planned — not yet built.Keys held as threshold shares — no single machine ever reconstructs them.
Planned — a design direction, not yet built.Encrypted-memory execution (SEV-SNP / TDX) on the hosted path, before the box exists.
Planned — a direction, not a scheduled release.Before the box
The operating system the hardware will run — a NixOS-based attested enclave OS whose whole image is a reproducible build you can verify byte-for-byte. It ships in the roadmap before the first box, because the box is only as trustworthy as the OS on it.
NixOS-based attested enclave OS with reproducible-build attestation — the verifiable software base the enclave hardware boots.
Planned — not yet built; sequenced before first hardware.On the horizon · Target H2 2027
The first sealed enclave hardware in owners' hands — the zero-fan, lid-watched box the whole architecture exists to earn, running MistOS. The pre-orders on this site are the signal that tells us when to pull the trigger.
A fanless, TPM-2.0, tamper-mesh sealed box — the first hardware, running MistOS v1.
Planned — a direction gated on the demand signal; owner pins Target H2 2027.Owner-run nodes that host apps and earn on idle.
Planned — a direction gated on the demand signal, with no committed ship date.Beyond the box
Once the box is in hands, it becomes a platform: a per-jurisdiction app economy with a revenue-share store, so builders ship enclave apps and owners earn from what runs on their node. The furthest horizon, gated on everything above.
A revenue-share store for apps that run inside the enclave, scoped per jurisdiction.
Planned — the furthest horizon, not a scheduled release.We build in the open and would rather be honest about what is real. If the direction resonates, come with us.
The vision, the engineering, and the wider set of use-cases — for builders, enterprises and investors.