Private Cloud Compute · All models · One place
.env → CEREBRAS_API_KEY for all users. This field is this-device BYOK. See config/COMPANY_CEREBRAS_KEY.example.
In the desktop app, phoenixd runs on this Mac — your seed, your node. This website is demo only. Click to open the wallet, or pop it out as its own rectangle.
Fund this device wallet, then spend sats on Cerebras inference from the AI tab. 100 sats ≈ 10 credits. Keys never leave this Mac.
Add USDB to the shared pool. Platform cut: 1.0%. The rest is shared immediately with all users.
Off is the default. Lightning (phoenixd), this Profile page, and the color controls stay fully clickable. Extra panels stay blurred until you turn Advanced Mode on.
Advanced Mode is on. Unblurred extras are now clickable but may not work until a later version. Sweep Lightning funds off this wallet before you download an update — this is not a hardware vault.
Whenever you install a new version, sweep funds first. Keys stay on this Mac.
No need to subscribe to multiple AI services. Top up as little or as much as you want. $1 = 1,000 credits.
The CLI lets you use Cerebras 3,000 tok/s inference directly from your terminal, Cursor, VS Code, or any CI script — authenticated via the same npub identity you created in the app. No extra accounts, no API key management. Your credits flow from the same wallet.
It will be a drop-in replacement for claude, openai, and llm CLI tools — same interface, orders-of-magnitude faster.
--pcc routes to Private Cloud Compute on supported Macs, 0 creditshm council "your prompt" queries all 3 LLMs, shows diffEach tier unlocks as you set things up.
higherMark_GridGlobal_0.1.5_aarch64.dmg · open the disk image → drag higherMark to Applications → right-click → Open the first time (Gatekeeper). Source: github.com/jimsbr/prodjuly3rdhM
Paste the Fourth Sol / higherMark company Cerebras key here. This is the pooled inference account. The same key belongs in .env as CEREBRAS_API_KEY (see config/COMPANY_CEREBRAS_KEY.example) so every user shares the pool.
Inference cost drops 0.05% for every 10 unique users (npubs). Formula: discount% = floor(users / 10) × 0.05.
We respond within 24 hours. Request new models, new workflows, report bugs, or share what you love.