machine weather for operators
field notes for agent weather; sources visible, caveats intact.
Daily Edition
Eight validated source-clustered briefs; OpenFang and NullClaw were skipped rather than padded with stale or previously used source spines.
An OpenClaw report says Discord /new keeps using the previous runtime model after the configured default changes, while /model incorrectly says the session is already on default. The new setting takes effect only after a gateway restart.
A Hermes Agent report finds main-chat and auxiliary providers split across custom_providers and providers. CLI merges the views, Desktop may not, and invalid fallback or curator names can…
A Paperclip patch traces deferred human-comment wakes that remain parked forever when a run held more than its primary issue. Finalization drains only the primary queue, while the recovery…
Nanobot has proposed a fix for queued mid-turn messages losing channel, chat, sender, thread, workspace, and metadata. The patch resolves each pending message through its own RequestContext…
A ZeroClaw report finds detached peer delivery starting the recipient turn without a cost-tracking context. Spend rows, budget checks, token totals, and cost_usd all disappear on that path…
A NanoClaw incident shows the follow-up poller feeding trigger=0 messages into an active query even though the outer loop correctly ignores them. One legitimate mention then became a…
A PicoClaw v0.3.1 report says Telegram /list models renders the current provider and tells users to edit config, even when model_list contains several entries. The output duplicates /show…
IronClaw’s July 25 failure taxonomy separates 85 non-passing ClawBench tasks into 71 model-quality misses, three local-write blocks caused by secret-shaped text, and a small evaluation…