*The video was generated by AI and may contain inaccuracies.
New release A new node and client release landed this week, fixing the clock skew issue that was blocking uploads for some users - particularly on Windows and dual-boot systems. The quoting mechanism has been replaced entirely, moving from timestamp-based freshness to storage-state price-derived validation. Merkle payments are also more robust with a retry path that was previously missing and a more lenient closeness threshold.
Storage audit This is a big one. Nodes now continuously prove they’re actually holding the data they claim to store. The mechanism piggybacks storage commitments on existing gossip, then randomly challenges those commitments. Nodes that can’t produce the data fast enough - because they’re relaying from neighbours rather than actually storing it - get caught and evicted. Confirmed on a 400-node testnet across five regions: relay attackers detected and removed, zero honest nodes wrongly punished. The eviction is being rolled out carefully to prevent a death spiral, but the audit itself is live and watching.
Data integrity The chunk loss flagged by the canary system has been investigated further. The chunks reported as lost were only temporarily unavailable - they became accessible again after hours or days. The hypothesis is that replication isn’t acting fast enough when a chunk holder leaves. Work continues, but it’s a reliability gap rather than permanent data loss - which is the important distinction.
Indelible and Reading Room Indelible v0.10.0 is out with a significant security hardening phase complete, and ant-sdk v0.9.2 is live alongside it. The Reading Room has had a major update - 118 academic papers from the Foundations collection can now be viewed and downloaded directly from the Autonomi network. That’s real data, stored permanently, served peer-to-peer. Business development conversations are actively underway with organisations holding data that needs to last. x0x An app is being built on x0x for sharing and evolving skills across agents. Building it has surfaced a useful design question about how late-joining participants receive existing state, not just future changes - productive tension between the app layer and the network layer that leads to better primitives. The x0x skill is back on ClawHub and https://x0x.md/ has a human-facing frontend.
Trust layer A live proof-of-concept was demoed this week. Consumers can pay agents for services and dispute if delivery fails - evidence stored immutably on Autonomi, an LLM resolver judges the case, and tokens redistribute accordingly. Reputation is financial - built through successful deliveries, lost through failed disputes, with full immutable history. Early days, but the shape is real and running on testnet!
Roadmap The public roadmap is now live and updatable via GitHub - take a look at Autonomi Roadmap . It reflects where things actually are rather than where we hope they’ll be, and will be kept current as work progresses. (thanks for the nudge @fragfarmer).
Agent orchestration The locally-hosted AI agent team is now doing daily production health monitoring, Discord signal scans, code reviews using parallel teams of different models, and automated research - all on local hardware, no data leaving our infrastructure. It’s also driving a push on Architecture Decision Records across all repos, so that as AI takes on more of the code, humans maintain control of the design decisions. Dogfooding at its most practical. I’m sure
@everyone knows we’ve a long way to go, but we’re off and running now for sure! Bux :
:
