Paying miners locked to Economy tier (0GB VRAM, 0 points) due to broken login

[Hafuyya]

Summary

Paying miner subscribers are completely locked out of earning points because hyperspace login fails server-side, which forces the node into "circular-economy" tier — Economy (CPU, 0GB) — regardless of hardware. This has been ongoing for over a week with no response from the team.

Impact

• Compute tier: Economy (CPU, 0GB) — despite RTX 5090 (32GB VRAM)
• Points: permanently stuck at 0.00
• Liveness multiplier: 0.00x
• Model inference: evicted after load (0GB VRAM budget)
• Points write mode: "crdt-only" (local only, never synced to server)

This affects all miners who cannot complete hyperspace login, which appears to be a widespread Supabase backend issue (see #27).

Root Cause

The tier evaluation in the CLI (v5.82.63) requires all three conditions to earn points:

evaluateTier() {
  if (this.airdropOptIn && this.isSupabaseRegistered && this.fullPulseVerification) {
    this.currentTier = "airdrop-qualified";
  } else {
    this.currentTier = "circular-economy";  // ← stuck here
  }
}

isSupabaseRegistered can only be true after a successful hyperspace login. Login fails with:

OAuth error: server_error
Callback: ?error=server_error&error_code=unexpected_failure&error_description=Database+error+saving+new+user

There is no env var, config option, or CLI flag to override the tier. The allocationMode config (power/balanced/chill) controls resource allocation but not the tier itself. HYPERSPACE_COMPUTE_ROLE only affects pipeline position, not tier. We have verified this by searching the binary.

Steps to Reproduce

1. Run hyperspace login → fails with "Database error saving new user"
2. Run hyperspace start --chain-role miner
3. Observe in logs: [ECONOMICS] Compute tier: Economy (CPU, 0GB) — mode: default
4. Model loads initially but gets evicted (0GB VRAM budget)
5. Points remain at 0.00 indefinitely

Environment

• CLI version: 5.82.63 (latest from download.hyper.space)
• Hardware: RTX 5090 32GB, Intel Ultra 9 285K, 64GB DDR5
• OS: Ubuntu 24.04, Linux 6.17.0
• Subscription: Active paid miner subscription (3+ months)
• Node: 17+ peers when networking is healthy, training sidecar functional

Additional Context

We have been paying for a miner subscription for 3 months. During this time we have filed issues #17 (sidecar reconnect crash, May 7), #18 (chain concurrent map crash, May 7), and #27 (login database error, June 5). None of these have received any response from the Hyperspace team — not even an acknowledgment.

• Issue Network sidecar crashes on reconnect: SidecarConfig required for NetworkNodeAdapter.start() #17: 35 days, 0 maintainer response
• Issue Chain binary crashes within seconds: concurrent map read/write in Sync.handleGenericConsensusMessage #18: 35 days, 0 maintainer response
• Issue Google sign-in fails with "Database error saving new user" (Supabase trigger failure) — blocks all new signups #27: 7 days, 0 maintainer response
• Last repo commit: April 17 (almost 2 months ago)

We've spent significant time debugging these issues ourselves — working around the port 4001 GossipSub self-conflict, the IPv6 EADDRNOTAVAIL crashes, the sidecar reconnect bug — but the auth/tier issue is entirely server-side and cannot be fixed from our end.

It is deeply frustrating to be paying for a service that is fundamentally broken, with zero communication from the team. At minimum, we need:

1. Fix the Supabase "Database error saving new user" bug so miners can log in
2. Or provide a workaround (API key auth, env var tier override, etc.) so paying subscribers aren't locked out
3. Acknowledge the open issues — silence for 5+ weeks on critical bugs is unacceptable for a paid service

Related: #27, #17, #18