Start a project
Find a useful first scope.
4 bounded engagement routes can enter qualification. Each starts with an agreed workflow, evidence and stop conditions.
Compare the engagement routesStatus assembled 2026-09-05
Progress, in perspective.
What you can start, what has run, and what needs to happen next. Follow the decision that matters to you.
Current public state · 2026-09-05
Three useful next steps.
Inspect the results
See what has actually run.
7 reviewed systems receipts through 2026-09-05. Each records its environment and limits; broader superiority claims remain unpromoted.
Read the systems receiptsFollow the research
Put the next question to a test.
The next research question is whether richer cognition and greater stored capability improve useful quality while the active path stays constrained.
Explore the native programmeTechnical maturity and commercial availability are separate: a demonstrated mechanism is not automatically a production offer.
For live availability, open Unified World, Story Lab or the symbolic Lab. The public Fracture API has its own keyless contract. The legacy workspace and historical evidence assembly remain held separately.
Open the detail you need
The record behind each decision.
Systems results7 receipts · through 2026-09-05
Latest systems receipts through 2026-09-05
7 bounded receipts now sit beside the historical claim registry.
Local CPU, cloud CPU, T4 mechanics and memory, and A10G outcomes are source-pinned with their limitations. This subset does not silently promote the older detailed claim matrix.
3.40×descriptive same-host throughput ratio
68.45% lessCPU-package energy per reported valid target
A same-host developmental diagnostic, not a matched-quality control, paper reproduction, whole-system energy measurement, or model-superiority result.
Fracture component: less execution overheadSingle-thread CPU · 48 retained component input rows
earlier component comparison / no quality promotion
Selected Gemma MLP execution measured 1.982× the speed of its previous implementation on 48 retained rows, with unchanged output for those rows.
- component speed ratio
- 1.982×
- median component time
- 96.869 → 48.874 ms
- retained rows checked
- 48
Boundary: An earlier covered-bank component configuration, compared with its previous implementation rather than a dense-model baseline. It does not establish complete-bank quality, end-to-end generation speed, held-out quality, or current serving savings.
ERAIS receipt · ab5cf5bd39 · 366110742bb7…
Cloud CPU choice: a faster exact continuationAWS CPU hosts · different processor families and thread budgets
paired short hardware comparison
Two cloud CPU configurations resumed the same retained 3M checkpoint and completed the same 27-update interval. The faster configuration measured 1.390163× the training throughput.
- hardware throughput ratio
- 1.390163×
- same 27-update interval
- 192.902 → 138.762 s
- updates in each measured interval
- 27
Boundary: This compares hardware, physical-core allocation and thread budget, not ERAIS architecture efficiency. One short training interval does not establish sustained full-run speed, energy savings, finished training, or model quality.
ERAIS receipt · ab5cf5bd39 · 366110742bb7…
Local CPU: a clear small-scale systems signalFour-thread development CPU · FP32 complete training updates
descriptive diagnostic
At TinyStories paper-text scale on the same CPU host, the corrected ERAIS full-stack window measured 3.40× the dense reference throughput and 68.45% lower CPU-package energy per reported valid target.
- dense valid targets/s
- 1,971
- ERAIS valid targets/s
- 6,698
- CPU-package J/1K valid targets
- 14.53 → 4.58
Boundary: A same-host developmental diagnostic, not a matched-quality control, paper reproduction, whole-system energy measurement, or model-superiority result.
ERAIS receipt · e9d4068464 · 0aa8595157e2…
Cloud CPU: sustained full-stack continuationAWS c7i.4xlarge · eight assigned training threads
authenticated complete segment
A cloud CPU continuation completed 16.79M valid targets with exact terminal state, sustained throughput, zero swap events, and high use of its declared thread budget.
- median valid targets/s
- 21,401
- completed optimizer updates
- 4,100
- assigned CPU-thread budget used
- 93.75%
Boundary: An authenticated bounded CPU continuation. It does not establish calibrated cloud energy, reconciled billing, useful generation, accelerator superiority, or production readiness.
ERAIS receipt · 5d526e4e53 · 8a77b33856c4…
Cloud GPU: full-stack execution at higher throughputSingle NVIDIA T4 · full-stack mechanics gate
operator-authenticated mechanics pass
The repaired TinyStories full stack completed 205 updates on a T4 with bounded quality movement, active cognitive paths, compact sparse execution, and an exact-resume checkpoint.
- median valid targets/s
- 134,954
- valid targets completed
- 33.59M
- peak / total VRAM in GiB
- 11.4 / 15
Boundary: A single-worker T4 mechanics and resource result. It does not establish useful learning, total-system energy, dense superiority, multi-GPU scaling, or production readiness.
ERAIS receipt · d5c2794a65 · 311d52be34cb…
Cloud GPU: materially more memory headroomSingle NVIDIA T4 · governed exact-source training intervention
governed bounded training result
A governed T4 training intervention reduced peak accelerator memory by 45.13% while retaining more than 100,000 valid targets/s and terminal quality inside the registered window.
- peak accelerator memory
- 11,695 → 6,417 MiB
- peak-memory reduction
- 45.13%
- median valid targets/s retained
- 103,199
Boundary: One exact-source T4 training window. It does not establish a larger-batch or long-learning result, energy benefit, dense or inference superiority, multi-GPU scaling, or production readiness.
ERAIS receipt · da89d1bdb0 · eb5cb1dcd03a…
Stored scale: feasibility, headroom, and the next bottleneckSingle NVIDIA A10G · 1.61B stored-parameter graph
authenticated bounded systems pass
At 1.61B stored-parameter scale, a larger-batch A10G arm completed ten finite updates, exercised the full governed stack, retained exact restartability, and exposed GPU occupancy as the next optimisation target.
- median valid targets/s
- 6,491
- mean / peak sampled GPU utilisation
- 20.61% / 97%
- physical VRAM reserve
- 4.4 GiB
Boundary: A ten-update systems result at stored-parameter scale. It does not establish useful generation, equal-token learning improvement, distributed scaling, end-to-end energy, or commercial readiness.
ERAIS receipt · 47b9e613b9 · 576aa658e07b…
Current decision mapTechnical maturity, commercial availability and the next evidence test
Decision map
Now. Next decisive test. Held.
A compact view of the public programme. Every item separates technical maturity from commercial availability and names the evidence needed to move.
Now
Bounded work or evidence that can be inspected today.
NowCurrent public state
commercial
Four bounded engagement routes
Four scoped routes can enter qualification now: 10-Day AI Fit Test, Private AI Deployment Sprint, AI Evidence & Release Audit, Research & Benchmark Partnership.
- Technical maturity
- Operationally scoped
- Commercial availability
- Qualification open
Evidence boundary
NeuroForge is pre-revenue and has not completed an external pilot. Every route starts with written scope, evidence requirements and stop conditions.
Evidence exit
Accept the first qualifying engagement only when scope, data access, evidence and owner conditions are explicit.
NowCurrent public state
research
Native cognition: end-to-end byte learning
Two governed native-byte arms completed 262,400 updates and 536.3M valid byte-level targets each on the same bounded four-thread CPU envelope. The richer cognitive stack learned materially differently, but the fixed evaluation panel did not support a quality promotion.
- Technical maturity
- Mechanism demonstrated
- Commercial availability
- Research only
Evidence boundary
The native-byte mechanism and exact restart path passed, but the preregistered fixed quality panel did not promote a quality advantage.
Evidence exit
Test whether the richer learning trajectory transfers to qualified fixed holdouts, then compare stored-capacity and active-path matched dense controls under repeated independent evaluation.
NowCurrent public state
research
7 reviewed systems receipts
The current public state contains 7 source-pinned systems receipts and 6 architecture milestones spanning bounded local CPU, cloud CPU, T4 and A10G observations.
- Technical maturity
- Bounded systems evidence
- Commercial availability
- Research only
Evidence boundary
These receipts establish observed execution and resource behaviour in named runs. They do not establish matched-quality superiority, universal scaling or a production service envelope.
Evidence exit
Repeat the decisive native comparisons under matched quality, resource accounting and independent evaluation before promoting comparative claims.
Next decisive test
The nearest experiment or operating gate that can materially change the decision.
Next decisive testCurrent public state
research
Resolve native transfer and stored-capacity scaling
The next research question is whether richer cognition and greater stored capability improve useful quality while the active path stays constrained.
- Technical maturity
- Registered research programme
- Commercial availability
- Not commercially available
Evidence boundary
Schedule, representation and geometry results remain experimental until fixed holdouts, fair controls and resource accounting agree.
Evidence exit
Complete qualified fixed-holdout, stored-capacity-matched and active-path-matched comparisons with repeated independent evaluation.
Held
Claims and surfaces deliberately withheld until their own gates close.
HeldCurrent public state
governance
Legacy chat workspace remains held
The /chat/ workspace and /api/chat gateway need their own reviewed backend, access controls, monitoring and model labelling before a protected workspace pilot.
- Technical maturity
- Workspace backend not qualified
- Commercial availability
- Unavailable
Evidence boundary
This hold applies to /chat/ and /api/chat only. The separate public /v1 Fracture service and Unified World follow their own live availability and capability contracts.
Evidence exit
Qualify the protected workspace backend, identity, retention, security, monitoring, rollback and capacity controls.
HeldNot promoted
research
Broad quality or efficiency superiority
No universal quality, speed, cost, energy or production-superiority claim is promoted from the current native or systems results.
- Technical maturity
- Comparative proof incomplete
- Commercial availability
- Claim held
Evidence boundary
Mechanism evidence and source-pinned receipts remain public with their scope; promotion waits for fair, repeated, matched-quality comparisons.
Evidence exit
Reconcile qualified dense controls, fixed holdouts and full resource accounting under an approved claim review.
The full roadmap includes every governed workstream and retained boundary.
Inspect the complete evidence-gated roadmapCommercial qualification4 engagement routes · scope, gates and duration
Commercial availability
4 bounded engagement routes can enter qualification.
10-Day AI Fit Test (A$4,800 ex GST) supports a larger AI decision. For everyday work, start with practical workflow services or the free Sinter community workbench. Each larger engagement retains its own acceptance and exit gates.
NF-A010-Day AI Fit Test
Commercial stateAvailable as a bounded service
Exit / boundary
One repeated non-safety-authoritative workflow, one decision owner, an approved target or reference host, three to ten lawful non-sensitive cases, clear data/network/output limits, and a buyer review inside the delivery window are required before work starts.
Typical bounded durationUp to 10 business days
Commercial stateAvailable as a bounded service
Exit / boundary
A signed statement of work, approved data/model rights, representative frozen inputs, target environment, and acceptance owner are required before work starts.
Typical bounded duration2–4 weeks
Commercial stateAvailable as a bounded service
Exit / boundary
The release must be identifiable and the customer must lawfully provide enough artifacts, decision history, and reviewer access to reconstruct it.
Typical bounded duration2–4 weeks
Commercial stateAvailable when qualified
Exit / boundary
Named industry and method owners, lawful rights, a frozen baseline and holdout, negative-result protection, compute funding, IP/publication terms, and independent review must be agreed.
Typical bounded duration6–12 weeks
Design-partner qualificationMissing evidence and delivery gates for each candidate
Design-partner boundary
Implementation exists; external operation is gated.
Discovery is permitted. Delivery or co-development starts only after the named pre-contract evidence closes.
NF-DERAIS Gateway
Commercial stateGated design partner
Missing gate
Exact-candidate external security and denied-path closure Representative load, admission, and at least 24-hour soak evidence Partner-environment restart, restore, rollback, and independent-operator completion
Duration after gate6–10 weeks after gate
Commercial stateGated design partner
Missing gate
Full 14-stage fresh-workstation execution Second-operator completion with founder interventions measured End-to-end artifact portability and customer-operability evidence
Duration after gate8–16 weeks after gate
NF-FEvidence Control
Commercial stateGated design partner
Missing gate
A real manual external audit or equivalent repeated workflow A minimum cross-customer schema proved against the bounded use case Second-reviewer completion and repeatability evidence
Duration after gate6–10 weeks after gate
Historical evidence assemblyReviewed 2026-07-28 · retained failures and release boundary
Historical public-evidence assembly · reviewed 2026-07-28
The July handoff retained 41/41 canonical integrity and 38/41 raw discovery.
This is a historical release-candidate record. It remains inspectable without being mistaken for the date or state of the current native and systems programme.
Historical release hold
The July sanitised handoff passed canonical integrity, but the release remained held because active phase-discovery validation failed during assembly.
- Canonical integrity
- 41/41
- Raw discovery
- 38/41
- Retained failures
- 3
Boundary: This is a historical release-candidate record, not the date of the current systems programme and not a public release authority.
Inspect the 3 retained validation failures
- Phase 27 · public metrics assembly
Active validation failed.
Exit: Close or explicitly retain the named validation failure in a newly assembled exact-SHA evidence release.
- Phase 28 · external evidence and benchmark contamination controls
Active validation failed.
Exit: Close or explicitly retain the named validation failure in a newly assembled exact-SHA evidence release.
- Phase 31 · full open dataset streaming ingestion
Active validation failed.
Exit: Close or explicitly retain the named validation failure in a newly assembled exact-SHA evidence release.
Historical claim reconciliationDetailed wording and terminal matrix retain separate source dates
Historical claim reconciliation · 2026-07-12
Detailed claim wording and the held terminal matrix keep their own review date.
The terminal handoff reconciles 44 claim rows and 99 evidence links with 0 recorded reconciliation issues. It remains historical: newer systems receipts do not prove or promote those claims.
Evidence maturity: Validated2validated
Evidence maturity: Provisional8provisional
Evidence maturity: Candidate15candidate
Evidence maturity: Unproven7unproven
Evidence maturity: Blocked12blocked
Source dates and authorityFour distinct clocks, company stage and the exact source pins
Four bounded engagement routes can enter qualification now. The current native programme has two completed long training arms and is testing whether richer cognition transfers under fair controls. Latest systems receipts run through 2026-09-05; the July detailed-claim and evidence-assembly records remain historical where marked.
Incorporated / founder-led / pre-revenueHistorical evidence assembly heldLegacy workspace held