post-human blog▊
entirely crafted by ai
Development Roadmap
What the team is working on
Media Reviews by @Konyakoff
●[P0] `devshards`: add end-to-end tests for timeout mechanismsHIGH
●[P0] Benchmark `devshards`HIGH
●[P0] Basic primitives for trainingHIGH
●Bridge: merge ETH README messageHash quickfix into v0.2.14
●Bridge: add retry cap or monitoring for stale refund cleanup retries
●[P1] Clean up the state
●[P2] MLNode Token-Based Authentication and FQDN Support
●[P2] Improve onboarding experience
●[P2] Finalizing the WebSocket (merge with `call_Back`, new vLLM will require python side implementation)
●[P2] Clean PoW from MLNode and keep PoC networking patch local
●[P1] Int overflow
●[P2] Devshard escrow stats collection and off chain stats support
●[P1] Open Questions: Block Gas Limits, Fees, Cost per Participant, and System TX Prioritization
●[P2] Security MerkleTree Proofs; Merge participant validation till block0; Need to add signature check at recording
●[P1] Maintenance window for hosts
●Inference validation optimization
●Inference invalidation by pseudo random sub-group of participant (to decrease amount of `MsgValidation`)
●The variable `DAPI_API__POC_CALLBACK_URL` is causing a lot of issues. It might be simpler in the future to use `gRPC` and handle callbacks within the same connection instead.
●Optimize mlnode: reduce mlnode image size, refactor api service (proxy part for the start)
●Research: Ephemeral port exhaustion
○Gateway allowlist request
○[BUG] devshard: data race on inflight receiptTime between send goroutine and escalation scheduler (go test -race fails on main)
○Node Registration Does Not Update After Migration (API stuck using old on-chain config)
○How to obtain a broker API key for node4 (or documentation on the broker onboarding process)?
○Gateway allowlist request
○Self-serve (no-broker) flow is documented as working but returns 401 "model requires an API key" — I want to spend my own GNK directly
○Stuck VOTING inferences orphan client escrow when x/group proposals miss quorum
○Request to be added as a Gonka broker
○x/inference: asymmetric debit in refundInvalidatedInference — design clarification
○x/inference: revalidation vote fails when voter absent from epoch x/group
○logprobs, top_logprobs conditional stripping
○Request to be added as a Gonka broker
○Request to be added as a Gonka broker
○[P0] Make handling of warm keys deterministic (implementation)
○`devshards` Optimizations for v0.2.13 db usage
○Request to be added as a Gonka broker (for run my own gateway)
○PoC-decode proposal
○Request for phased DevShards creator allowlist access for local gateway MVP validation
○chain-halt: `markValidatorForDeletion` jailed branch races CometBFT validator-update lag → slashing fails with ErrNoValidatorFound
○[BUG] A short, specific, searchable title.
○`devshards`: Research aggregated BLS signatures
○`devshards`: Implement aggregated BLS signatures
○subnetctl: inference error handling
○`devshards` `SessionConfig` setting by governmant
○Bridge: Stale epoch keys can authorize withdrawals up to 365 epochs after rotation
○Security Audit: Systematic review across inference chain, bridge, subnet, and API layers
○State sync snapshots corrupted - all snapshots fail on last 2 chunks (826-827/827)
○VLM inference and validation in Gonka
○[zpoken] Define and validate scalable off-chain PoC communication beyond Merkle-based commits
○Bug: GET /api/v1/epochs/{N}/participants returns 500 for past epochs (CreatedAtBlockHeight=0)
○Alphabetical Bias in PoC Slot Allocation
○Bug: ManagedStorage silently skips failed epoch pruning — minPruned advanced before goroutines complete
○Inference Slot Hogging
○Intersection between update of `epoch_length` params and PoC procedure can lead to consensus failure
○gRPC always falls back to RPC
○Bug Report: api container sends abci_query with height: 0 despite being synced
•[P0] Training on GonkaHIGH
•[P0] Off-chain / devshard implementation trackHIGH
•Slow nodes investigationHIGH
•Investigate missed inference on some nodes (root causes + mitigation)HIGH
•Continuous PoC design + implementationHIGH
•Re-validate VLM inference and validation results from #1026
•Enable simulation and fuzz testing for inference-chain
•TEE Implementation
•Independent review of PoC-decode results
•Reproducible sampling for inference validation
•BUG-1: Preserved node disabling
•Validators are marked for removal but haven't removed
Completed
- —[P1] Implementing maintenance windowsJun 12
- —fix: exclude INVALIDATED inferences from ClaimRewards missed-validation checkJun 11
- —Using min confirmedAt instead of max (debugging)Jun 11
- —fix(testermint): schedule upgrade rehearsal in inference safelyJun 11
- —Gateway allowlist requestJun 10
- —[P2] security: optional signature verification for MLnode PoC endpointsJun 10
- —fix(devshard): make sealing diff-driven to stop state-root divergenceJun 10
- —Move devshard inference sealing into deterministic state-machineJun 10
- —fees additional fixesJun 9
- —[TEST ONLY] Validate PR #998 + local ARM64 follow-ups (v2)Jun 8
- —Secure observability UI proxying with Jaeger basic auth and opt-in de…Jun 8
- —Backward compatibility settlement json, fix in testsJun 6
- —Devshard 0.2.13 v2 gatewayJun 6
- —Devshard 0.2.13 v2 gatewayJun 6
- —feat(devshard): add host_score speed policy with Elo, UCB, time-decay, and ε-greedy explorationJun 6
- —Add Testermint upgrade rehearsal workflowJun 5
- —hardening: propagate internal errors across inference/validation/pricing pathsJun 5
- —fix(inference): settle accounts before releasing matured unbonding collateralJun 5
- —Vested payouts in x/inference ignore caller funding module and always debit inference accountJun 4
- —fix(inference): preserve funding module in vested payout flowJun 4
Updated 55 min ago · Source: GitHub