Run, then fix
Generated code runs in a sandbox. If it fails, the error goes back to the model and it tries again. Python, JavaScript, Bash, PowerShell, C, Brainfuck.
Spectral HX: incorporation in progress
Chuck, the IDE super-coder.
Spectral HX puts Chuck inside a full desktop IDE. It runs what it writes, fixes what fails, and remembers the solutions that actually worked. Underneath sits an emulated anyonic woven-fabric compute layer: a classical simulation of braided anyons that Chuck uses as a structured compute core, with the language model as the interface, not the intelligence.
No card required to try. Capability ceiling: not yet measured. The free Chuck on osv01d.org stays free.
What it actually does
Shipped capabilities, not roadmap items.
Verified-solution memory
Code that exits zero is stored and matched against future work. The assistant gets better at your problems.
Local-first inference
A quantised model runs on your hardware. Offline, no per-token cost, no prompt leaves the machine unless you pick a cloud model.
Visible routing
Every request shows which expert pocket was selected, at what weight, and whether the admit gate accepted it.
Consent-gated system probe
Chuck asks before reading toolchain versions, every time, and stores only version tokens, never paths or usernames.
Topological lab
Bell and GHZ circuits on a four-node Majorana ring, Fibonacci and Ising braid compiling. Real simulation, clearly labelled as simulation.
The woven-fabric compute core
English in, structured job out, results back as English, visuals, sound or files.
- Translate. The language model turns your request into a structured job.
- Weave. The job runs on an emulated 32×32 anyonic fabric, a classical braiding simulation.
- Verify. Results are executed and checked before they reach you.
- Render. Chuck explains the result in plain English.
No quantum speedup is claimed. The fabric is emulated on ordinary hardware.
Pricing
Founding pricing is live and locked for 24 months. Full-price plans unlock only when the capability gate below is met.
Founding Personal
$49.99/month
- 1 seat
- Local inference
- Run-and-fix loop
- Local verified memory
Founding Pro
$79.99/month
- 3 seats
- Everything in Personal
- Verified-memory sync
- Priority routing
Pro
$199.99/month
- 5 seats
- Shared memory bank
- Basic SSO
- Priority support
Enterprise
$299.99/month
- Unlimited seats
- SAML / OIDC
- Audit log
- 4-hour response SLA
Annual billing gives two months free on every tier. Personal and Developer list-price plans ($99.99 and $149.99) open at the same gate.
Status, stated plainly
Where we're ahead
Privacy, offline operation, verified memory and visible routing.
Where we're behind
Raw code correctness and generation speed still trail cloud models. That's why list pricing is gated.
The gate
Full pricing opens after sustained high correctness, faster generation, and 30 days without a billing or licence defect.
Support that answers like a person
A named human signs every ticket. Published response targets: 48h Personal, 24h Developer, 12h Pro, 4h Enterprise. Cancel in one click.