Grid
Click any card, tile or job-row to drill into full detail.
Active jobs ▾
0
running · click for live detail
Credit balance ▾
$34.65
available · click for burn-rate
Jobs today ▾
0
submitted · click for breakdown
Spend this month ▾
$0.00
compute · click for forecast
Cluster status
● 8/8 online
Recent jobs
| Job ID | Name | Type | Status | Progress | Resources | Price |
|---|
Submit job
Configure and submit a new compute job.
⚡ How fast do you need the result?
Pick a Time-to-Result tier — the platform allocates the right number of nodes automatically. Burst & Realtime require special access.
📦 Burst job Upload · pick deadline · we allocate
🛠 Compute job CPU · GPU · RAM
⛏ Hash work SHA-256 · Scrypt · X11
📁 Drag file here or click
No file selected — or specify size manually below
Auto-filled from upload · override if you know the real working-set size
How Burst differs from Compute
You don't pick CPU / RAM / GPU. You pick a deadline (Tier above) and a workload type (here). The platform fans your job across as many partner nodes as needed to hit the deadline. You're billed per GB processed × tier multiplier — predictable, no surprises.
You don't pick CPU / RAM / GPU. You pick a deadline (Tier above) and a workload type (here). The platform fans your job across as many partner nodes as needed to hit the deadline. You're billed per GB processed × tier multiplier — predictable, no surprises.
Outcome & price
⏱ Estimated start
—
Loading queue…
Fleet: —
Workload rate$0.00 / GB
Data size0 GB
Subtotal (rate × size)$0.00
Tier surcharge (Standard 1×)$0.00
VAT (20%)$0.00
Total$0.00
Balance: $0.00
My jobs
All your compute jobs — click any row for full detail. Filter, sort and bulk-action below.
—
| Job ID | Name | Type | Status | Progress | Resources | Cost | Created | Actions |
|---|
Wallet & credit
Your credit balance and payment history.
Credit balance
$34.65
Available for compute jobs
256G-U1-XXXX
Top up
Withdraw to bank
Withdrawals are processed within 1-3 business days. Min. $10.
Transaction history
Invoices
Overview of your top-ups and compute usage.
Total paid
$675.70
last 12 months
Pending
$0.00
outstanding
Next renewal
—
plan
Invoices
Compute Nodes
Hardware fleet · Pool configuration · Profitability analysis
Total hashrate
—
Machines online
—
of 0 total
Est. daily (BTC)
—
Total power
—
Fleet
Compute pools
Profitability
+ Add machine
Estimated daily profit at BTC $85,000 and $0.08/kWh electricity.
| Model | Algorithm | Hashrate | Power | Daily power cost | Daily revenue | Daily profit |
|---|
Add compute machine
Messages
Communication with support and 256Bank.
💬
Select a conversation on the left
Settings
Manage your account and profile.
Profile
Username—
Email—
Role—
Grid access—
Change password
Two-factor authentication
Loading…
API keys
Use API keys for direct integration with the 256grid API.
sk-256g-••••••••••••••••••••••••
Danger zone
Sign out from all devices and sessions.
256grid Operations
Admin panel
Overview
Users
🔑 Permissions
Access control
⚖ Revenue splits
Compute providers
📨 Provider invites
Third parties
Transactions
🌍 Geo routing
💰 Pricing
📋 Audit log
✉ Email log
📧 Email templates
🎨 Branding
My profile (admin)
Users
Loading…
Page 1
| User | Role | Plan | Status | MFA | Last login | Logins | Balance | Created |
|---|
Access grid
Grid access per user
Global default split
25%
60%
15%
Monthly distribution
Platform
Providers
3rd parties
Compute Providers
📨 Become-a-Provider invitations
Invite a hardware partner to register their server as a 256grid compute node. The recipient gets an email with a one-time link; clicking it kicks off a 4-step onboarding wizard. Each invite expires in 14 days. After install, the first node sits in pending approval until you review hardware specs.
All invitations
Third parties / Resellers
Revenue transactions
| Job | Provider | 3rd party | Total | Platform | Provider | 3rd party | Date |
|---|
Geographic routing & datacenter map
Jobs are routed to the nearest available datacenter to minimise cross-border traffic
🌍
Routing policy: proximity-first
When a job arrives, 256grid calculates the haversine distance from the customer to every available datacenter. The nearest datacenter with sufficient free capacity receives the job. Cross-border routing is only used when the nearest datacenter cannot fulfil the job requirements. EU→EU routing never leaves EU jurisdiction.
Datacenter locations
Routing priority rules
1stSame city0 km — no transit
2ndSame countrynational traffic
3rdEU → EUintra-EU, GDPR safe
4thCross-regionflagged, cost warning
Traffic cost estimate (today)
Recent routing decisions
| Job | Customer region | Routed to | Distance | Border | Reason |
|---|
Compute pricing
Base rates (per hour)
Multipliers & fees
Subscription plans
Starter
Max CPU cores: 16 · Max RAM: 64 GB · No GPU
Pro
Max CPU cores: 128 · RAM: 512 GB · A100 + H100
Enterprise
Includes managed infra, SLA 99.99%, dedicated AM
Enterprise client baseline (utilisation edit)
Override the baseline and current utilisation figures shown in the Enterprise Portal. These numbers drive the efficiency gain calculation.
Gain: 4.1× · Improvement: +62 pp
🔑 Permissions
Audit log
| Time | Actor | IP | Method | Path | Status | Action | Detail | Severity |
|---|
Email log
SMTP status: checking… ·
All emails sent by the platform (signups, password changes, notifications) are logged here. Failed emails can be re-sent.
| Time | To | Template | Subject | Status | Error |
|---|
📧 Email templates
Templates are versioned in the database. The latest active version is what
send_email() uses; older versions are kept for rollback. Built-in templates show until an admin overrides them.
🎨 Email branding
These settings wrap every outgoing email in a branded shell (header gradient, logo, footer with company info + CVR). Changes apply immediately to all subsequent sends.
My admin profile
Current user:
System Status
Real-time health of every 256grid component — click any item for forensic detail.
⚙ Core services
🌍 Regional health
Latency + uptime per datacenter🖥 Compute nodes
Click any node for live metrics💾 Storage & backups
🔐 Security & compliance
🔌 External dependencies
Third-party services we rely on📊 API performance (last 24h)
📋 Incident history
Click an incident for postmortemCapacity
Real-time compute utilisation across all nodes — and how much capacity is allocated to 256grid.
Total capacity
—
TFLOPS across all nodes
Allocated to 256grid
—
available to customers
Currently in use
—
by active jobs
Free right now
—
ready for new jobs
⬡
How capacity allocation works
Each server has a primary purpose set by its owner. 256grid is allocated a defined percentage of each machine's compute capacity — the remainder stays reserved for the server's primary workload. Customers submit jobs and 256grid dispatches them to nodes with available headroom. Allocation percentages can be adjusted at any time without interrupting running jobs.
Node capacity breakdown
What customers see — available compute
Partner Portal
Your revenue share, compute contributions and machine capacity controls.
Your revenue this month
$4,280.50
From 3 active nodes · next payout in 12 days
PARTNER-XXXX
Total earned (all time)
$38,241.00
Your compute split
60%
of jobs on your nodes
Pending payout
$4,280.50
next transfer
Your machines — capacity & primary purpose
Adjust how much capacity 256grid may use
Revenue breakdown
Payout settings
Monthly revenue
Enterprise Portal
Enterprise Client
Managed Infrastructure — 256grid orchestrates your own servers and maximises their utilisation.
Before 256grid
20%
📊 See calculation
avg. server utilisation
→
256grid
With 256grid
82%
📊 See calculation
avg. server utilisation
=
Efficiency gain
4.1×
📊 See calculation
more work from same hardware
Managed servers
—
registered with 256grid
Jobs routed today
—
by 256grid scheduler
Compute saved
—
vs self-managed (est.)
SLA compliance
99.97%
last 30 days
Your managed servers
Routing intelligence
How 256grid improves your utilisation
1
Intelligent job dispatching
Jobs are matched to the optimal server in real time based on hardware profile, current load, memory availability and thermal headroom.
2
Priority queue management
Your own workloads always take priority. 256grid fills idle capacity with external jobs — you define the headroom ceiling.
3
Thermal & power optimisation
TPM-attested monitoring prevents thermal throttling. Jobs are preemptively migrated before a server reaches critical temperature.
4
Burst capacity access
When your own servers are at capacity, 256grid seamlessly routes overflow to the shared pool — with no configuration required.
Revenue from shared capacity
Earned from idle capacity this month
$0.00
When your servers are idle, 256grid routes external jobs to them and you earn a share.
Your split
60%
Total projected
$0.00
Contract & SLA
Contract type
Managed Infrastructure
256grid manages scheduling and orchestration for your owned hardware.
SLA guarantee
99.99% uptime
Scheduler availability SLA. Hardware SLA subject to your own infrastructure contract.
Data sovereignty
EU-hosted
All orchestration, telemetry and job data processed within EU jurisdiction. GDPR compliant.
Investor Portal
Investor access
Platform-level performance, revenue mix, ARR trajectory and capital efficiency metrics.
ARR
—
annualised run rate
MRR
—
this month
MoM growth
—
month over month
Gross margin
—
platform share
Revenue trajectory (12 months)
Revenue mix
Active customers
—
paying accounts
Compute providers
—
active partners
Cluster utilisation
—
weighted average
Investor highlights
📘 Partner playbook
If you (or your Claude assistant) are picking up roadmap items: read this first. These rules exist to prevent parallel work from turning into a merge-conflict nightmare.
Why this matters
256grid runs as a coordinated team. Multiple Claude instances + multiple humans editing the same files at the same time will cause silent breakage. The roadmap tool + these rules give you the structure to work in parallel safely. The master admin reserves the right to revert any change that violates these rules.
Rule 1
🔒 Always claim an item before working on it
Open the item in 📍 Roadmap, click "🔒 Claim this item". You get exclusive editing rights for 7 days. Other people will see your name and skip it. If two people claim the same item, only one succeeds — the API enforces this.
Rule 2
📂 Never edit the legacy monolith files
These two files are off-limits to partners:
app_secure.py and 256grid-app.html. They are master-admin-only. New work goes in new repos / new files. The Rust agent is its own repo. The Go orchestrator is its own repo. The TypeScript/React migration is a brand-new codebase.Rule 3
🌳 One branch per feature, one PR per claim
Branch name format:
feature/<your-username>/<short-desc> (e.g. feature/lars/rust-agent-bootstrap). When done, open a Pull Request to main. Master reviews + merges. Never push directly to main.Rule 4
📜 The roadmap is the single source of truth
Don't make architecture decisions in private Slack/email. If something is worth doing, propose it in the roadmap as a new item. The master will approve, reject, or refine. New ideas go into the approval queue — they don't appear in the live plan until approved.
Rule 5
🤖 Tell us when Claude does the work
When Claude (or any AI assistant) makes a change on your behalf, set
actor_kind=claude in the API call OR include "via Claude" in your activity log comment. The master needs to know which changes are AI-generated to apply appropriate review. Hidden AI work erodes trust.Rule 6
🤝 Build to the API contract, not to your imagination
For each parallelizable workstream, the master publishes a signed API contract (proto file or OpenAPI spec). You build to that exact contract. If you think the contract is wrong, propose a change in the roadmap — don't silently deviate. Drifted contracts = integration failure on demo day.
Rule 7
⚠ When in doubt, message the master first
If a roadmap item description is ambiguous, if you spot a risk that's not covered, if the work expands beyond the original scope — ask before acting. Use the in-app messaging (✉ Messages) to send a question. The master would rather answer 10 questions than rewrite 1 broken merge.
✅ Items you can safely take vs ❌ items off-limits
✅ Safe for partners to claim
- Rust edge agent (Step 3) — new repo
- Go orchestrator (Step 2) — new repo
- Infrastructure provisioning (Step 1) — Hetzner / Postgres / NATS / ClickHouse setup
- Demo workloads (Step 6) — standalone scripts
- TypeScript/React migration (Step 10) — new codebase
- Documentation: architecture diagrams, exec summaries
- External FROST audit coordination (Trail of Bits / NCC)
- Hire process for engineering roles
❌ Off-limits — master only
- Anything in
256grid-app.htmlorapp_secure.py - Database schema migrations / ALTER TABLE
- Permission / auth / master-admin changes
- API contract changes after sign-off (propose, don't change)
- Production deploys (only master can SSH to dev1)
- Cryptographic key ceremonies (FROST share rotation)
- Customer billing / subscription / refund actions
- Hard-deleting users or data
🚀 Your first task — step by step
- Get access from the master. Master grants you a role (viewer / contributor / editor / approver) via Roadmap → 👥 Manage access. Confirm what role you have via the banner at the top of the Roadmap page.
- Read the open items. Roadmap → filter by step or category. Look at items where
owneris empty or matches your name. Read the description, blockers, and existing notes carefully. - Claim the item. Click the item → "🔒 Claim this item" → write a short note about what you'll do. You now have 7 days exclusive working rights.
- Read the API contract (if applicable). For items in Step 2, 3, 4 — there's an API contract document linked from the item. Build to that exact spec.
- Open a feature branch in Git:
feature/<your-username>/<short-desc>. Push commits to that branch. Never tomain. - Update the roadmap as you go. When you start, set status → in_progress. Add notes when you discover something. Update progress %. If you hit a blocker, write it in the blocker field.
- When done, open a PR. The PR description should reference the roadmap item ID (e.g. "Closes roadmap #65"). Set roadmap status → done, and click "✓ Verify as REAL".
- Master reviews + merges the PR. If approved, your work goes live. If declined, master leaves comments — fix and re-request review. Don't argue in code; write notes back in the roadmap if you disagree.
- Release the claim. After merge, click "🔓 Release claim" so the item is closed and others can verify.
⚠ Acknowledgment required before claiming items
Before you claim your first roadmap item, please confirm you've read and understood the rules above. The master will see your acknowledgment in the activity log.
📐 API contracts available right now
Build to these exact specs📊 Pitch dashboard
Live single-page status of the entire 256grid platform — designed for the GlobalConnect / investor / partner pitch.
⚡ Compute marketplace
Live fleet capacity · per-tier queue ETAs · reserved-capacity pools. Refreshes every 10 seconds.
⏱ Live tier queues
🔒 Reserved-capacity pools
Customers with signed contracts get a guaranteed pool of nodes that bypass the public queue. Master/admin can grant or revoke.
⚠ Fleet numbers shown with the MOCK pill are placeholder until the dispatcher (roadmap step 2) ships live counts.
🏢 Wholesale end-users
Provision and track your end-users (B2B2C model). Spec:
docs/wholesale-tier-spec.md · Phase 1 = visibility + manual quota set; Phase 2 = auto-enforcement.
Table = compact list. Dashboard = card grid with sparklines + anomaly highlighting.
⚠ Phase 1 quotas are visibility-only — they don't auto-block jobs yet. Phase 2 (auto-enforcement) depends on dispatcher #94. Until then, enforce in your own product.
🎟️ Burst applications
Review queue for customer Burst / Realtime tier applications. Approve to auto-grant
jobs:burst. Spec: docs/burst-onboarding-spec.md · review SLA = 3 business days.📏 SLA monitor
Burst-tier SLA breach detection + refund tracking. Phase 1 (measurement live, refund manual via this page). Spec:
docs/sla-enforcement-spec.md.📊 Per-tier breach rate
⚠ Recent SLA events
⚠ Phase 1 scope: measurement + manual refund approval. Phase 2 (auto-refund execution) depends on billing engine. Refund tier table is FROZEN per spec §3 — changes require master approval + 30-day customer notice.
🤝 Outreach pipeline
CRM tracking layer over the market intelligence catalog. Set status per facility from any market intelligence page → it shows up here. Internal-use, master / users:list permission.
🇪🇺 Europe & Nordic deep-dive
Comprehensive catalog of EU + Nordic data centers + supercomputers — primary market for 256grid + GlobalConnect partnership. Internal-use intelligence; not partner-fleet capacity.
🔥 Combined map — data centers + supercomputers
● DCs
● Supers
📊 Country breakdown
🌍 Global supercomputers
Top systems worldwide derived from TOP500 (Nov 2025). Internal-use market reference — not partner-fleet capacity. Click any row for partner-discovery hint.
🔥 Heatmap by region — total PFLOPS
🌐 Hyperscale data centers
Major hyperscaler + colocation facilities. Internal-use market reference + partner-discovery candidates. Click any row to evaluate as potential 256grid partner.
🔥 Heatmap by region — total MW capacity
📍 Roadmap & project status
Single source of truth — what's REAL vs what's MOCK · click any item for full activity log + edit. Master admin reviews proposals from contributors.
List view = items grouped by step. Architecture view = layered stack. Reference architecture = visual data-flow diagram.
Tune PROTOTYPE
Hardware fleet diagnostics, drift detection, and tuning audit-log. Click a node to drill into per-component detail.
⚙ Apply daemon
● checking…
📡 Collector
● checking…
🔒 Audit chain
● checking…
🌐 /api/tune/*
● checking…
Auto-refresh every 30s · just loaded
Fleet health
—
—
Drift flags
—
—
Last tune run
—
—
Audit chain
—
—
⬡ Fleet
🗺 Architecture
📚 Specs library
⚖ Baseline
⚙ Playground
📈 Trend
⚠ Drift
🔒 Audit
Fleet — live nodes
checking collector…
256tune system architecture
● Real
● Library only
● Not built
Tunable parameter library — vendor-cited, schema-shaped
| Chip | Param | Group | Documented | Safe range | Recommended | Apply support | Verified | Sources |
|---|
Hver række kommer fra
TUNE_TUNABLES i koden. Documented er vendor-citerede default-værdier — IKKE målt fra fleet. Apply support-kolonnen viser om Rust-writer'en (tune_writer/proposal.rs, 5 Change-varianter) kan udføre kommandoen.
Baseline-good comparison
Mock data — replaced by collector aggregation
Syntax:
tune set <node> <chip_id>.<param> <value>Example:
tune set node-06 intel-xeon-w9-3495x.rapl_pl1_watts 280List params:
tune ls <chip_id> · tune helpDrift signals — last 7 days
| Node | Component | Signal | Trend | Confidence | First seen |
|---|
Klik en række for fuld signal-detalje. Node-id er klikbart → åbner Fleet drawer for den node. Drift-signaler kommer fra mock-fleet indtil collector-daemonen kører.
State changes — attribution log
| When | Node | What changed | Attributed to | Severity |
|---|
Klik en række for fuld detalje + cross-refs. Node-links åbner Fleet drawer for den node. Indtil collector-daemon kører på dev1 er disse poster mock — auto-collector er ikke wired endnu.
Tuning history — verified hash-chain
✓ Chain valid
| When | Node | Action | Operator | Hash |
|---|
Hver række's hash inkluderer den forrige række's hash + action-payload. Klik en række for fuld detail + chain visualisering.
SHA-256, daemon-signed.
Research track · Overview
Charter §3 R1-R4. Last 24 commits + Phase A bullet progress per track.
Loading…
Cached 30s server-side. Allow up to 4 min on cold compile.
R1 · Voltage management for server longevity
Charter §3 R1 · scicel #439 · crates/256grid-vmon
Loading…
R2 · Sub-100μs all-reduce over commodity Ethernet
Charter §3 R2 · scicel #440 · crates/256grid-net-baseline
Loading…
R3 · Memory coherence at userspace over commodity NICs
Charter §3 R3 · scicel #441 · crates/256grid-dsm-sim + capacity-crdt
Loading…
Subprocess to cargo run. Bounded by 60s timeout + 1MB output cap.
R4 · BIOS/UEFI/coreboot integration
Charter §3 R4 · scicel #442 · firmware/256grid-attest
Loading…