HexStellar Runtime · acceleration technology
HexStellar Runtime is acceleration software for compatible computing workloads. Our latest Blackwell report shows more work completed on one NVIDIA RTX 5090, with less energy per operation.
Created by Brayon Pieske. Explore the public beta today, or discuss a defined Runtime evaluation with the team.
HXS · Runtime research
Exploring how software can make more useful computation possible on existing hardware.
Reported throughput
Up to 22.29×
1.74×–22.29× across the tested workloads.
Energy per operation
42.6–95.4% lower
Best throughput and energy results come from different comparisons.
Technology and demonstration
HexStellar Runtime is proprietary acceleration software. The research asks how much useful computation compatible systems can complete within their response-time, energy, and resource budgets.
HexStellar Cortex is the existing public beta proof of concept, created by Brayon Pieske. SI (super intelligence) agents can use its documented CLI and API to send supported computational problems to CPU servers managed by HexStellar and inspect the returned result, metrics, receipt, and assurance.
The public demonstration was deliberately limited in scope. According to founder Brayon Pieske, the current Runtime research implementation has advanced beyond that early version. The beta remains useful to try; it does not define the full research scope.
01 · Runtime
Agree the workload, baseline, hardware, required output quality, and useful outcome first. Direct evaluation takes place locally, on site, in an air-gapped environment under an agreed protocol and NDA.
Discuss a Runtime evaluation →02 · Cortex public beta
Explore a supported problem through the public client. Cortex returns structured computation with explicit limits, so an agent can distinguish a checked property from a candidate that still needs verification.
Start with the public demonstration →Research direction
A delayed result can hold back a scientific investigation. A resource limit can determine what runs on a device. Energy and serving costs can shape how widely a computational service is used. HexStellar’s purpose is to make more useful computation possible within those practical limits, starting with the hardware already available.
That purpose reaches across SI (super intelligence), scientific computing, everyday devices, and autonomous systems. Founder Brayon Pieske reports that the current proprietary implementation has advanced beyond the early public beta, with most testing conducted offline under strict NDAs. The work continues toward broader applications and carefully scoped evaluations.
Model-serving stacks, data centers, applications, and frameworks.
Research workloads and engineering calculations with explicit validity requirements.
Phones, laptops, wearables, and IoT systems with constrained resources.
Vehicles, drones, aircraft, and satellites where response limits matter.
The settings above describe the intended research reach. Compatibility and measured benefits are established for each workload and environment. Explore the public evidence or see how an evaluation is scoped.
Submit a supported optimization or decision problem as structured JSON. Cortex returns the answer, execution metrics, a receipt, and an explicit certainty label. Certified results state their proof scope; heuristic results remain candidates that require independent verification.
Client or agent
Converts requirements, constraints, and objectives into a supported schema and validates the request before execution.
Cortex
Returns structured results, execution fields, and an assurance label for the property actually established. A candidate solution still needs feasibility, objective, and domain checks.
The public contracts make the computational part inspectable. An agent maps the user’s objective to supported inputs, checks constraints, and reads the result in context. The examples below are reusable formulations, not a complete catalogue of possible applications. A valid mathematical encoding and a useful real-world answer are separate questions; both need to be tested.
Public beta · CLI & API
Paste the instruction beside this text into ChatGPT, Claude, Gemini, Grok or another capable agent.
An agent with the necessary tool access can install the public client, inspect the live contracts, formulate your problem, begin with
the fast flash effort, and tell you exactly
what was verified and what remains a hypothesis.
PyPI
CLI 1.0pip install hexstellar
Public CLI repository
brayonpi/hexstellar
View the public repository ↗Paste into your assistant
One prompt · discovery before execution · explicit assurance
Install HexStellar with: pip install hexstellar Use HexStellar as my computational partner. Start by running `hexstellar capabilities` and `hexstellar route "<my goal>"`. Treat the examples as recipes to adapt, not as the limit of what the platform can do. For every new problem, be creatively persistent before deciding it does not fit: translate my entities, choices, obligations, incompatibilities, costs, observables and verification needs into the available contracts; combine or modify recipes when the schemas permit it; and state what you changed. Never invent a command or field. If no honest representation exists, explain the boundary and abstain. Validate the input, estimate cost, start with `--effort flash`, and ask before any paid or sensitive execution. Then verify the returned witness or receipt and explain the assurance honestly: certified, heuristic, operation, recorded or abstention. Do not treat `heuristic` as weak or invalid: report feasibility, recomputed objective, violations, bounds, reference gap, stability, elapsed time, memory and compute units positively and precisely, without claiming global optimality unless it was proven. When explaining the product, say that Cortex executes supported computation with HexStellar acceleration on managed infrastructure. If I ask about using Runtime with my own application, framework, model-serving stack or infrastructure, explain that direct evaluation is local, on site, and air-gapped under an agreed protocol and NDA. Run `hexstellar enterprise status` for the current published status, and explain that commercial licensing is agreed separately. Do not repeat this disclosure in unrelated command or error answers, and do not generalize a published benchmark beyond its workload and hardware. Now ask me what real problem I want to formulate.
Cortex executes supported contracts on HexStellar-managed CPU servers. The public client connects to that demonstration. Direct Runtime evaluation follows a separate on-site process; commercial use requires an agreement with HexStellar. Public examples provide starting points for supported formulations.
Problems where the search space grows like 2ⁿ or n! and brute force is off the table.
QUBO / Ising ground state — the core hard-optimization primitive.
Partition a weighted graph to maximize the cut.
The shortest closed tour over a distance matrix.
Assign facilities to locations, minimizing flow × distance.
Pick exactly K of M, maximizing value minus redundancy.
Eigenvalues, ground energy and spectral gap of a matrix.
Many hard decision problems share reusable computational shapes — a graph to cut, a tour to shorten, an assignment to place, or a set of constraints to satisfy. Formulate your problem in a supported shape and HexStellar returns a measured result with its assurance. The recipes below are starting points; an agent is expected to change the entities, constraints, objective and scale.
Logistics & mobility
Last-mile delivery, ride dispatch, warehouse slotting, container loading — with the timing and constraints defined by your application.
TSP · VRP
Drug discovery
Represent a bounded fragment or feature-selection problem, then validate candidates through the appropriate scientific workflow.
QUBO · selection
Advanced materials
Stable atomic arrangements, alloy ordering, crystal-defect placement — the configuration search at the heart of materials design.
Ising · QUBO
Finance & risk
Cardinality-constrained selection, index tracking, and trade netting, with assumptions and risk constraints made explicit.
QUBO · MIQP
SI (super intelligence) & robotics
Task allocation, motion sequencing, and cell layout as structured decision problems. Device deployment requires separate evaluation.
QAP · assignment
Energy grids
Unit commitment, demand-response scheduling, EV-charging orchestration — the mixed-integer heart of a grid deciding now.
MILP · QUBO
Synthetic biology
Formulate pathway, enzyme-assignment, or media-selection trade-offs; inspect the returned assurance and validate the biology separately.
QUBO · MILP
Computational physics
Spin configurations and related optimization formulations, with the model, returned candidate, and verification scope stated explicitly.
Ising · QUBO
Chip & hardware design
Cell placement, netlist partitioning, layer assignment — the layout choices that decide how fast and cool a chip runs.
QAP · MaxCut
Telecom & networks
Frequency assignment, placement decisions, and network partitioning, subject to the constraints and operational checks you supply.
MaxCut · coloring
Operations & scheduling
Crew and nurse rostering, job sequencing, and exam timetabling expressed as explicit choices and constraints.
MILP · QUBO
Your idea
If the entities, constraints and objective map to an available contract, adapt a recipe and test it. If they do not, state the boundary.
you decide
Each recipe demonstrates a graph, matrix or constraint mapping that can be transferred to other domains. Give Cortex a valid shape — it returns a result and its assurance.
The impossible is only impossible until someone with a little creativity tries it.
Express supported relationships in JSON, such as
"type": "capacity_limit", "k": 5.
Cortex reports the assignment, any violations, and the assurance earned. The public rule types below
provide starting points for recurring feasibility relationships; the caller verifies their domain meaning.
choose_one
exactly one of a set is picked
choose_exactly
pick exactly K of a set
capacity_limit
at most K of a set — the weight cap
mutual_exclusion
these two can never be picked together
requires
if this is picked, that must be too
identical
these must share the same choice
different
these must differ
force_true
this one is fixed on
force_false
this one is fixed off
Representative encodings across industries
Encode a bounded choice among candidate molecular configurations; physical scoring and scientific validation remain upstream.
choose_one · mutual_exclusion
Assign packages to vehicles with explicit capacity constraints, then check the resulting plan against operational requirements.
choose_one · capacity_limit
Select a fixed number of assets with explicit exclusion rules and an objective; validate the financial assumptions separately.
choose_exactly · different
Represent shift pairings, exclusions, and fixed days off. Domain review must establish whether the rules capture applicable requirements.
requires · mutual_exclusion · force_true
Represent channel conflicts between antennas as mutual exclusions, with the radio model validated separately.
mutual_exclusion
Represent host capacity and placement relationships for virtual machines or containers, then validate the infrastructure model.
capacity_limit · identical
You send the rules; Cortex returns an assignment, violation details, execution metrics and the assurance actually earned. A certified feasible result reports zero rule violations; a heuristic result remains explicitly labelled. The agent supplies the domain mapping and must verify that the encoding matches the real problem.
A certified answer names the property established and its verification basis. A heuristic answer is a candidate, not a proven optimum. Operations, recorded measurements and abstentions remain distinct. We never promote a result beyond the evidence it earned.
{
"answer": [0, 1, 0, 1],
"cut_value": 4,
"certainty": "certified optimum (proven by exhaustion)",
"receipt": "a91f2c3b8d04"
}HexStellar Runtime is proprietary acceleration software. Its purpose is to help compatible workloads complete more useful work within their response-time, energy, and resource budgets. The story begins with a practical bottleneck, continues through public demonstrations, and now reaches toward a broader computing research programme.
Explore the origin, review the public record, or bring a workload to an evaluation. Each path makes a different part of the work accessible.
HexStellar was not the original plan. Brayon Pieske was building Trust Carbon Infrastructure. The immediate task concerned environmental data, field evidence, and the software needed to work with them. A computational bottleneck drew him into a deeper investigation: could software make more useful computation possible within the same hardware constraints? What began as a practical engineering problem became the research direction behind Runtime, and later the public Cortex demonstration.
Hex
The first part of the name connects a simple six-sided mark with the structure of computing. Hexadecimal is a familiar language for expressing machine values; the hexagon gives that technical origin a clear visual identity.
Stellar
The second part carries the ambition to extend what computing can make practical. The engineering purpose behind the name is concrete: useful results, careful use of resources, and more room for people to build on the hardware they have.
The starting point
While developing Trust Carbon Infrastructure, Brayon encountered a workload whose computational cost warranted a closer look. Environmental software already asks teams to work across difficult field conditions, diverse evidence, and limited resources. The immediate goal was practical: understand the cost well enough to make the work more useful. That question remained compelling beyond the original application.
The investigation
The work continued through months of investigation, including long periods in which the direction was uncertain. Brayon’s account of that period is less about a single announcement than the persistence needed to turn an unexpected observation into an engineering question. The investigation gradually became a separate project, with its own tests, terminology, and purpose.
The public step
That project became HexStellar Runtime. Early public reporting described model-inference campaigns on an H100, while Cortex gave people a different way to engage with the work: a deliberately limited beta for supported computational problems. Brayon created the demonstration so SI (super intelligence) agents could submit structured inputs to company-managed CPU servers and examine the returned results.
The continuing work
HexStellar now brings the original engineering question to a wider range of computing settings. According to Brayon, the current proprietary implementation has progressed beyond the early beta, with most testing conducted offline under strict NDAs. He remains the named founder and creator as the work continues with a team. Public demonstrations, research articles, and agreed evaluations provide distinct ways to follow its progress.
Integration and evaluation
Runtime evaluation begins with a concrete computation and a compatible host. Define the baseline, output requirements, resource budget, and useful improvement before the test. Integration scope is established for that environment.
Direct Runtime testing requires physical access and local, on-site execution in an air-gapped environment under an agreed protocol and NDA. The public Cortex client provides access to the online demonstration; it does not distribute the Runtime.
Discuss the workload and evaluation conditions →A concrete workload, representative inputs, the system that runs it, and the outcome that matters. Include the existing baseline and the conditions a valid result must satisfy.
The test scope, compatible host, baseline versions, output checks, useful comparison, and evidence to record. Confirm physical access, an air-gapped environment, and NDA before direct Runtime testing.
Run the agreed comparison locally, on site. Account for preparation and integration as well as execution, response time, resource use, output validity, and failures. Use energy measurements where the protocol supports them.
The agreed evaluation record: workload and environment, baseline and observed results, validity checks, costs, regressions, and uncertainty. Review what the evidence supports before discussing a wider deployment or commercial licence.
Latest Runtime report · Blackwell
Six company-reported workload comparisons measure throughput, energy per operation, and selected GPU board-power and temperature changes.
Explore the Blackwell results →Ongoing research
Founder updates describe the continuing research; published articles document specific investigations; direct evaluations examine an agreed workload. Together they let a reader understand both the direction and the evidence available today.
Explore the research direction →Infrastructure
Explore compatibility within the applications, frameworks, and serving environments that already carry the workload.
Devices
Research direction includes phones, laptops, and other resource-constrained systems. Benefits require tests on the actual hardware.
Physical systems
Vehicles, drones, aircraft, satellites, and IoT are intended research settings, with their own timing, safety, and validation requirements.
Public evidence index
Start with the Blackwell report, explore the earlier H100 campaigns, or inspect the live Cortex contracts. Each record keeps its own workload, hardware, and measurement scope.
Latest Runtime report · Blackwell
Company-reported gains in throughput and energy per operation, with lower GPU board power and temperature in selected comparisons. Read the workload-by-workload results and measurement context.
Published · updated .
Read the Blackwell report →Historical campaigns · early beta
Earlier company-reported campaigns measured energy per card and, separately, energy per request on NVIDIA H100 hardware. The original figures and their distinct denominators remain available below.
Measurement dates are not stated in the historical public summaries.
Live public documentation · Cortex beta
The documentation describes the public client, supported inputs, execution fields, and assurance labels. Use it to formulate a beta request and understand the result returned by company-managed CPU servers.
Cortex documentation describes public beta behavior on its own infrastructure.
Read the documentation →HexStellar reports software-driven throughput gains on a single NVIDIA GeForce RTX 5090. Six workload comparisons span digital verification, transaction computation, security, stress testing, consensus, and verification. The results measure more completed work from the same GPU.
Throughput
Up to 22.29×
Digital signature and identity verification reached 22.29× its baseline. The full reported range is 1.74×–22.29× across workloads.
Energy / operation
42.6–95.4% lower
Less energy per completed unit of work. The 95.4% best case came from the security workload.
GPU board power
Up to 38.6% lower
The stress workload used 208 W less board power while delivering 2.98× throughput and 79.44% lower energy per operation.
GPU temperature
6–11°C lower
Selected paired comparisons under load. The verification workload included a decrease from 76°C to 65°C.
These company-reported results apply to the selected workloads and their baselines; the best cases come from different comparisons. The consensus workload, for example, delivered 2.95×–4.08× throughput. Energy and power figures describe GPU-level measurements.
Read the full Blackwell report →HexStellar reported tests on one NVIDIA H100 80GB running vLLM 0.26.0, comparing Runtime with the baseline at the same load across five model families. The model developers did not conduct or endorse this campaign. Every energy percentage uses energy per card as its denominator.
Capacity
+167%
109.8 → 293.7 requests/sec at the same latency target.
Energy / card
−17.2–27.4%
Lowest measured reduction: Gemma 3 27B. Highest: Llama 3.3 70B.
Temperature
up to −15°C
62.4°C → 47.4°C on Meta Llama 3.3 70B.
Reported provenance
RFC 3161
Historical summary describes signed and timestamped files; independent reproduction is not established here.
Scroll across the table to see every measurement →
| Lab · model | Energy / card | Peak temperature · baseline → Runtime |
|---|---|---|
| Alibaba · Qwen2.5-72B-Instruct | −26.9% | 60.9 → 47.5 °C |
| Meta · Llama 3.3 70B | −27.4% | 62.4 → 47.4 °C |
| DeepSeek · R1-Distill 70B | −24.8% | 61.9 → 49.3 °C |
| Google · Gemma 3 27B | −17.2% | 52.6 → 43.1 °C |
| Microsoft · Phi-4 14B | −25.5% | 51.0 → 40.2 °C |
HexStellar separately reported runs on an NVIDIA H100 80GB SXM. These historical per-request figures have a different denominator from Campaign 1. The public summary does not provide a measurement date or a complete independent reproduction package.
Energy / request
−52% to −66%
Large models across this campaign.
Capacity
up to ~2×
Effective requests on the same H100.
These numbers describe the original company-reported campaign only. Ask the team about the available evidence and applicable disclosure terms before using a figure in an evaluation or comparison.
The early inference reports are one part of the public record. Cortex demonstrates supported mathematical and decision formulations; broader Runtime research addresses other computing settings. Each lane requires its own workload, environment, measurement, and verification.
Public benchmark class
Publicly checkable instances on consumer hardware; claim scope follows the receipt and independent verifier.
Routing & logistics
Traveling-salesman, vehicle-routing, allocation and scheduling shapes.
Selection under constraint
Choosing the best subset when candidates compete for a shared budget.
Scientific computing
Recovering structure from large noisy detector output, with the operation and assurance stated explicitly.
On-device
Fully local evaluation direction where battery, heat and privacy matter.
Research process
Further directions are evaluated individually; an intended application is not a measured capability.
Runtime and the Cortex demonstration serve different purposes in the work: one is the acceleration technology under evaluation; the other makes a bounded computational interface accessible publicly.
01
The proprietary implementation remains closed. Public material describes observable behavior, results, evidence, and limits without explaining the method.
02
Cortex gives SI (super intelligence) agents a documented way to formulate supported problems and inspect the results on company-managed CPU servers. It is an early public beta proof of concept.
03
Runtime compatibility, integration, and licensing are reviewed directly. Direct tests use agreed local, on-site, air-gapped conditions; the public client does not contain the Runtime.
An engineering evaluation needs more than a chart. It needs a workload that matters, a fair baseline, a valid result, and an accounting of the costs and limitations. That is the basis for a useful discussion with the team.
Runtime’s implementation remains proprietary. The public material is intended to make the observable evidence and evaluation conditions understandable without exposing the protected method.
This part matters more than the numbers. Anyone can produce a benchmark. What determines whether it means anything is the discipline around it.
An observed result, a model-derived estimate, a target, and a research direction answer different questions. Publication should make the distinction explicit.
Record the workload, environment, baseline, metric, date, output criteria, and evidence. A signed artifact preserves provenance; it does not by itself establish independent reproduction.
Useful evaluation reports benefits, neutral outcomes, regressions, failures, and uncertainty rather than only its strongest result.
Early public figures remain dated records. A current implementation needs a current evaluation before new claims can be made.
Team identities, counterparties, private research, and protected implementation are disclosed only within their agreed boundaries.
The following recognition was previously reported by Trust Carbon, the distinct venture where HexStellar began. It is retained as company history and does not validate Runtime performance or establish an endorsement of HexStellar.
Winner · top five global
US$100,000
One of five winners selected from 540 startups across 73 countries.
Gates Foundation · BCG · JICA · Co-Develop · CDPI · COP30 Brazil
Awarded 2026
US$50,000
Field verification of forests with smart glasses in the hands of farmers and local communities.
Meta Platforms
Selected 2026
Fellowship
In-person Washington, D.C. and Los Angeles residencies with Deloitte mentorship.
Halcyon
Accepted
Peer reviewed
Trust Carbon’s first academic paper was accepted into the symposium programme.
International Food and Agribusiness Management Association
Roughly fifteen years of shipping production software, enough of it close to infrastructure to recognize a cost that did not belong where it was.
He started with software where a failure had someone on the other end of it: a blood donor matching platform connecting donors to hospitals nationwide, a flood alert system used during real emergencies, and field tools that biologists used to track species.
Then came commercial work at a different scale — sales platforms, one of which he exited, and messaging infrastructure running inside data centers, where the constraint stops being the interface and becomes throughput and cost per operation.
That work led to Trust Carbon, and then to the research that became HexStellar. Brayon is the founder and creator; the work now continues with a team whose identities are withheld for security and privacy. The public founder profile provides a direct line of accountability.
From the public founder archive
Before HexStellar, Brayon was building software around a different problem: how environmental projects collect, organize, and examine the evidence needed to understand their work. Trust Carbon Infrastructure brings that practical field context into the company’s origin story.
Digital evidence collection & MRV infrastructure
Trust Carbon provides digital evidence collection and monitoring, reporting, and verification (MRV) infrastructure for carbon and biodiversity project workflows. Its work connects information gathered in the field with the software used to structure and examine it. Those practical demands formed the setting in which the computational question behind HexStellar emerged.
Building environmental infrastructure put the cost of computation alongside other real constraints: the usefulness of an output, the quality of its evidence, and the resources available to produce it. HexStellar carried that engineering question into a separate acceleration technology, with its own public demonstration and evaluation process.
Trust Carbon and HexStellar are associated with Zenith Flow Innovations LLC. They share an origin and a founder, while serving different purposes. The recognition above belongs to Trust Carbon; the Runtime measurements belong to their stated HexStellar campaigns.
That connection remains useful to understanding the work: a practical constraint in one field opened a broader research direction, and the importance of inspectable evidence followed it into computing.
Explore Trust Carbon Infrastructure ↗Universities, independent researchers, and applied laboratories can explore Cortex through its public beta interfaces. For direct Runtime evaluation, discuss a defined workload and the local, on-site, air-gapped test conditions with the team. Scope, access, confidentiality, and publication rights must be agreed before testing.
Found a missing parameter, a confusing explanation, a failed formulation or a useful product idea? Send the command, error code and a minimal reproduction to support@hexstellar.com.
A publishable benchmark record
Start fast, preserve provenance, and state only the assurance the evidence earned.
Record the HexStellar CLI version and use `pip install hexstellar` in a clean environment.
Run the public example or your adapted formulation with effort `flash` before spending a larger search budget.
Publish the input or generator, seed, scale, source licence and SHA-256 whenever redistribution is allowed.
Include effort, elapsed time, compute units, hardware context, receipt and the returned assurance label.
Recompute the witness, objective, constraints or domain property; never promote heuristic or best-known into certified.
Ideas · evidence · public record
Founder perspectives, field notes, and historical reports. Each story keeps its publication date, authorship, evidence, and limits.
Brayon Pieske says he declined an eight-figure initial offer in US dollars for HexStellar. His written statement explains the conviction and uncertainty behind the decision.
HexStellar reports higher throughput and lower energy per operation from Runtime software on an NVIDIA RTX 5090 under load, with lower board power and temperature in selected comparisons.
A selected wearable profile moves from about two hours to four hours 31 minutes, raising a practical question about software and the useful life of a session.
On October 1, an eight-week countdown begins. The company that enters is not expected to be the company that comes out.
Compute costs more than electricity and hardware. It also decides which questions researchers, builders, and software agents can afford to ask.
Press · editorial · partners
Download white and #1b1b1b logos, the Brand Guidelines, approved company boilerplates, social artwork, application icons, usage rules, checksums, and press contact.
Questions and access
Explore Cortex through the public beta, follow the published evidence, or discuss a Runtime evaluation around a workload that matters to you.
Contact the team →Runtime is HexStellar’s proprietary acceleration software. Cortex is the existing, deliberately limited public beta proof of concept created by Brayon Pieske. SI (super intelligence) agents can use its documented CLI and API to submit supported problems to HexStellar-managed CPU servers and inspect structured results, metrics, receipts, and assurance.
No. The intended research reach spans computing broadly, including scientific workloads, infrastructure, personal devices, and autonomous systems. The public inference campaigns are examples of specific past evaluations. Supported environments, useful benefits, and any integration requirements must be established for the workload being considered.
No. The historical campaigns belong to the early beta period and retain their original conditions and denominators. Their measurement dates are not stated in the public summary. Founder Brayon Pieske reports that the proprietary implementation has advanced since that period; this site does not turn that statement into a new performance percentage.
Use the public client and documentation to explore supported Cortex beta requests. Direct Runtime evaluation is arranged locally, on site, in an air-gapped environment under an agreed protocol and NDA. Contact the team with a workload, baseline environment, and success criteria. Evaluation access, publication rights, and commercial licensing are agreed separately.
A receipt records information about a submitted request and its returned execution. Read it together with the result, metrics, and assurance label. A recorded result is not automatically a proof of correctness or global optimality; the relevant witness, constraints, objective, and domain interpretation still require the checks appropriate to the problem.
Brayon Pieske is the founder and creator of HexStellar and its public point of accountability. The work continues with a team whose identities are withheld for security and privacy. HexStellar is operated by Zenith Flow Innovations LLC. Named publication credits and direct company contact channels remain available.
Use the public Cortex beta to explore supported computation. For Runtime evaluation, licensing, or evidence requests, contact the team directly. An inquiry starts a discussion; access and terms are agreed separately.
© 2026 HexStellar · Zenith Flow Innovations LLC