Scalar · Proof of AI infrastructure
Useful intelligence should be computed, verified and shared.
Scalar is a distributed AI-compute and validation network being developed by Phronesis Labs. It coordinates useful computational work, verifies the results through independent nodes and creates an auditable intelligence record that can serve the NeuroTrader ecosystem and other approved systems.
Tokenless testnet in development
Proof of AI · POAI
Proof that useful intelligence was actually produced.
Proof of AI is Scalar’s proposed protocol for assigning computational work, recording how it was performed and independently validating the result. Nodes receive credit for work that satisfies defined requirements — not simply for consuming electricity or returning an answer.
Proposed protocol · Testnet specification in development
Useful work
Jobs must serve an approved computational, AI, research, simulation or validation purpose.
Signed execution
Worker nodes return signed records identifying the job, process, model version, inputs, outputs and completion evidence.
Independent validation
Separate validator nodes reproduce, compare or challenge the submitted result.
Weighted agreement
Acceptance can consider validator agreement, evidence quality, operator reliability and job-specific thresholds.
Auditable provenance
Every accepted result retains a traceable record from job creation through final validation.
POAI is designed to evidence how a result was produced and validated. It does not guarantee truth or accuracy, and it does not eliminate AI error.
System flow
Jobs enter, work is proven, intelligence is absorbed.
01 · Submit
Phronesis Labs or an authorized external system submits a bounded job with permissions, constraints and acceptance criteria.
02 · Decompose
The job is divided into isolated work packets that can be assigned without exposing unnecessary data or unrelated customer context.
03 · Mine
Worker nodes perform useful computation. In Scalar, “mining” means completing approved AI or computational work — not solving meaningless puzzles.
04 · Prove
Each worker produces an output and a signed work receipt containing the required execution evidence.
05 · Validate
Independent validators reproduce, compare, score or challenge the work.
06 · Braid
The job record, worker proof, validator attestations, disputes and final decision are bound into one tamper-evident intelligence lineage.
07 · Absorb
Accepted results enter the Scalar intelligence layer and can be delivered to the requesting application. Rejected or disputed results remain quarantined.
Stages describe the intended testnet architecture. Components not yet operational are labelled as planned.
The Braid
Every intelligence result carries its history with it.
The Braid is Scalar’s proposed provenance structure. It binds the original job, computational output and independent validation into one traceable record.
Job strand
- Who submitted the job
- What was requested
- Permitted data and tools
- Success criteria
- Job version and timing
Work strand
- Assigned worker node
- Execution method
- Model or process version
- Input and output hashes
- Completion evidence
Validation strand
- Validator identities
- Reproduction and comparison results
- Agreement or disagreement
- Challenges and resolutions
- Final acceptance status
The Braid makes it possible to inspect not only an answer, but where it came from, how it was produced and why the network accepted it.
Network roles
Four responsibilities, kept apart.
Useful-work miner
- Completes assigned AI or computational work
- Operates within defined resource and permission limits
- Returns signed outputs and evidence
- Builds reliability based on accepted work
Validator
- Independently checks miner output
- Reproduces or challenges results
- Detects inconsistent, malformed or manipulated work
- Signs an acceptance, rejection or dispute
Coordinator
Planned component- Routes work during the controlled testnet
- Selects appropriate miners and validators
- Prevents collisions and unsafe assignments
- Monitors reliability and network health
Cortex
Planned component- Combines accepted intelligence
- Resolves validated outputs into usable context
- Delivers results to authorized applications
- Feeds observed outcomes back into future assignment and validation
The intelligence loop
Validated work supports decisions. Outcomes improve the network.
Scalar is designed as a learning infrastructure loop. Validated work supports decisions. Observed outcomes improve future job routing, validator weighting and intelligence quality.
- Request
- Compute
- Validate
- Braid
- Absorb
- Apply
- Measure outcome
- Improve↻
Private customer information is not used for automatic model training. Outcome measurement operates on network routing and validation quality.
Inside the ecosystem
One intelligence network. Multiple specialized applications.
NeuroTrader
- Market research and decision support
- Strategy evaluation
- Backtesting and simulation
- Decision History analysis
- Risk and execution-context validation
EventEdge
Coming soon- Event evidence processing
- Probability comparison
- Catalyst monitoring
- Countercase and outcome validation
Perpetuals
Coming soon- Funding and positioning analysis
- Liquidation and risk simulations
- Market-regime evaluation
- Multi-source validation
NeuroTV
Coming soon- Research compilation
- Evidence verification
- Programme and briefing generation
- Traceable source lineage
Connected AI
Coming soon- Authorized ChatGPT, Claude, Grok or other supported systems may request Scalar jobs
- Scalar can return validated results and provenance
- Connected assistants do not receive unrestricted customer information
- Every connection remains permissioned and revocable
Beyond our ecosystem
Validated AI infrastructure for other systems.
Scalar is also being designed as infrastructure that approved companies, applications and research systems can access through APIs and software-development tools.
Planned · Not yet commercially available
Potential external uses
- AI-output verification
- Research validation
- Model comparison
- Distributed simulations
- Data-quality checks
- Backtesting workloads
- Reproducibility audits
- Provenance and signed work records
Planned external flow
- 01External system
- 02Scalar API
- 03Permissioned job
- 04Worker nodes
- 05Validators
- 06Braid receipt
- 07Verified result
External jobs remain isolated by customer, workload and permission boundary. Results are not contributed to collective intelligence unless the submitting party explicitly authorizes it.
Discuss an external integration
Tell us about the workload. We will follow up privately with the technical review process — no capability is presented as commercially available.
Scalar on devices
Distributed intelligence running where the network lives.
Scalar node device
A preconfigured dedicated device for approved worker or validator participation.
Phronesis One / NeuroTrader device
An ecosystem device capable of running NeuroTrader applications and, when enabled, contributing bounded background compute to Scalar.
Third-party computer
An approved computer or server running the Scalar node client within defined hardware, security and connectivity requirements.
Planned node software
- Scalar node runtime
- Secure node identity
- Sandboxed workload execution
- Signed work receipts
- Validator client where authorized
- Health, contribution and reliability dashboard
- Resource controls
- Automatic updates and rollback
- Pause and disconnect controls
- Encrypted connectivity
Operator controls
- Scalar participation is opt-in.
- Operators can see when work is running.
- Operators can set resource limits.
- Nodes cannot access unrelated customer information.
- Devices can be paused or disconnected.
- Battery, thermal and connectivity limits must be respected.
Selected testnet operators may be offered preconfigured hardware at no upfront cost when the device program opens. Selection, availability, ownership, return requirements and operator obligations will be disclosed before distribution.
Decentralization roadmap
From a controlled testnet toward a distributed protocol.
Phase 1 · Controlled tokenless testnet
- Company-coordinated job routing
- Permissioned miners and validators
- Direct payment for approved services when activated
- Security and validation testing
- No token
Phase 2 · Federated network
- Independent worker and validator operators
- Multi-validator agreement
- Published POAI and Braid specifications
- External API pilots
- Direct monetary compensation
Phase 3 · Distributed protocol
- Broader independent operation
- Reduced reliance on one coordinator
- Transparent network and reputation rules
- Distributed validation
- Governance architecture evaluated
Phase 4 · Possible network token
- Under legal and technical evaluation
- No commitment to issue
- No guaranteed distribution
- No guaranteed exchange availability
- No rights created by early participation
Node network registration
Help test the network.
This is a public testnet-interest list for miners, validators, developers, research contributors, integration partners and device-program applicants.
Future network financing
Potential Scalar SAFT under legal structuring.
Phronesis Labs is evaluating a possible Simple Agreement for Future Tokens to support Scalar’s development. The structure, exemption, eligibility, terms, timing and jurisdictions have not been determined.
No SAFT or token is currently being offered
Request future SAFT information
This request is separate from the node network list. We do not ask for a proposed investment amount, payment information, a wallet address, accreditation documents, a commitment to purchase or a token reservation.
Nothing on this page is an offer to sell or a solicitation of an offer to buy any security or token. No money or other consideration is being requested or accepted, and if sent it will not be accepted. Joining this list creates no obligation, commitment, allocation or eligibility of any kind for either party.