Attention shows who gets seen; verified updates show which facts changed and which evidence boundaries still need confirmation.
Dates use UTC. The default view shows the newest event date; when an older event enters later, its recorded date appears separately.
How external evidence enters public history
A submission stays private until its source is checked, its public record is edited bilingually, and the patch is accepted.
See how evidence patches workChange type
DAILY VERIFICATION RECORD
BitShovel organized “the AI says it is done—how do I know it is deliverable?” into understanding, comparison, evidence, a verification card, and practice.
Why it matters
Products no longer stand as isolated profiles; they return to a reviewable problem and decision process.
Evidence boundary
This proves the page and checklist are public; it does not prove they have helped real users ship.
The work is split into content publication, hands-on product validation, and user impact; only publication is marked complete.
Why it matters
“Live” is no longer expanded into “the product works” or “growth happened,” so the next action can target the missing evidence.
Evidence boundary
BitShovel has not installed AgentLimb, completed a real-account task, or demonstrated user impact.
DAILY VERIFICATION RECORD
The profile checks the official site, Chrome Web Store, public repository, and privacy policy, using official demo material.
Why it matters
It adds a concrete approach for observing real browser-agent actions and permission boundaries.
Evidence boundary
This is a public-source review; the extension and local Bridge were not installed, no account was authorized, and maker efficiency claims were not reproduced.
The profile checks JetBrains' product page, public-preview announcement, Windows release notes, and official interface images.
Why it matters
It provides a comparable example of an agent control plane built around plan review, parallel execution, and final diff inspection.
Evidence boundary
The product remains recorded as a public preview; BitShovel did not install it, sign in, run an agent, or validate output quality.
DAILY VERIFICATION RECORD
The US App Store now shows version 1.0.17 at 177.3 MB; the profile also retains the mismatch between website and store ghost-count descriptions.
Why it matters
A recheck updates mutable facts while preserving unresolved differences between sources.
Evidence boundary
BitShovel did not install or purchase the app on iPhone or validate sensor, performance, network, or privacy claims.
DAILY VERIFICATION RECORD
The profile adds the Microsoft Store Windows listing and three product images from the Apple App Store; those store facts do not independently verify agent behavior or data paths.
Why it matters
Public store listings can establish platform and media facts without turning maker presentation into runtime verification.
Evidence boundary
BitShovel did not install the app or independently audit network behavior or interception coverage.
The profile adds the publicly available first-party technical article and code-flow context while retaining a “partial verification” conclusion.
Why it matters
More implementation detail narrows the unknowns, but a real form and final PDF are still needed to validate the result.
Evidence boundary
BitShovel did not independently complete a real form, inspect the final PDF, or audit runtime behavior.
DAILY VERIFICATION RECORD
The profile checks the official site, transparency and feature pages, pricing, privacy, terms, extension, and MCP documentation.
Why it matters
It puts the public capabilities, data paths, and authoritative-source boundaries of a domain-research tool in one profile.
Evidence boundary
BitShovel did not query a domain, register, install, connect MCP, call the API, monitor, or pay.
The profile checks the official site, API, MCP, SDK, pricing, privacy, terms, and public machine-readable manifests.
Why it matters
It turns API and MCP side effects, data handling, and version boundaries into inspectable questions.
Evidence boundary
BitShovel did not register, request a key, submit content, connect MCP, install an SDK, test rate limits, or pay.
The profile checks product architecture, pricing, FAQ, privacy, download endpoints, and official interface images while retaining a public version mismatch.
Why it matters
It places pre-delivery risk checks for AI-generated code, data payloads, and version evidence in one decision frame.
Evidence boundary
BitShovel did not download, install, authorize, enter a key, monitor network traffic, or validate detection coverage.
DAILY VERIFICATION RECORD
The profile checks the open-source repository, license, configuration, compatibility matrix, security notes, and cloud-data boundaries.
Why it matters
It adds the evidence layer for tracing actual agent runs, fixing evaluation criteria, and comparing outcomes.
Evidence boundary
Only a read-only public repository check was performed; BitShovel did not install, run, create an account, or send a trace.
The profile checks the task board, terminal state, live diffs, history, permissions, pricing, and roadmap.
Why it matters
It provides a control-plane example for supervising parallel agents while making clear that auto-approval is not security isolation.
Evidence boundary
BitShovel did not install the product, run an agent, or test Turbo Mode permission enforcement.
The profile checks the public editor demo, pricing, GitHub permissions, privacy, security, temporary collaboration, and retention notes.
Why it matters
It puts pre-execution specifications and review in the same collaboration context as post-execution diffs.
Evidence boundary
BitShovel did not sign in, authorize a repository, or independently validate permission and merge behavior.
Follow public changes via RSS
Updates, problems, solutions, and practices share one subscribable timeline.
