MIKE — The Demo Nobody's Watching Carefully Enough
Mike OSS shipped on 8 May 2026. It demonstrates — cleanly and honestly — that open-source tooling can generate a legal artefact from a prompt. What the demo does not show is the hour that follows: who receives the artefact, what they do with it, who is accountable for the outcome. That hour is the gap Donna is built to close.
What Mike actually shows
Mike is Will Chen's open-source legal-AI tool. The demo is well-constructed: a lawyer enters a prompt, Mike produces a document. The generation quality is credible. The latency is low. It is a genuine demonstration of what local, self-hosted inference can now do for legal document work.
That is exactly what the demo shows. It is worth watching carefully for a second reason: what it does not show.
The demo ends when the document is produced. It does not show the document being routed to a reviewer. It does not show the reviewer being briefed on client context. It does not show the completed review being confirmed and filed against the matter. It does not show who is responsible if the review does not happen.
These are not features Mike omitted carelessly. They are not in Mike's scope. Mike is a document-generation tool. Generating the document is the correct scope for Mike. The gap after the document is a different layer entirely.
The artefact/coordination gap
The manifesto's competitive table places Mike, Harvey, and Legora in the same column: they generate the artefact. Donna sits in a different column: it coordinates the work around the artefact (manifesto §7).
This is not a criticism of Mike. It is a structural observation about where the two tools sit. A Delegation Orchestration Layer does not compete with a document-generation tool. It sits above it. A lawyer who uses Mike to draft and Donna to coordinate the work around that draft has closed both sides of the gap.
The demo is worth watching carefully because it shows the gap precisely. Every minute of coordination work that begins after the Mike demo ends — who reviews this, by when, on whose authority, with what confirmation — is the minute Donna is designed to absorb. Law 1 of the manifesto (Orchestration Over Features) and Law 4 (Delegation Proof Over Task Storage) exist to address exactly what comes after the document appears.
Open-source legal AI is not a threat to a Delegation Orchestration Layer. It is the clearest argument for one. The better the generation tools get, the more visible the coordination gap becomes — because the artefact arrives faster, and the surrounding work does not.
Falsifier. This framing is wrong if Mike's roadmap includes a coordination and delegation layer in a near-term release — i.e. if Mike moves from artefact-generation into orchestrating the work around the artefact. Check Mike's public roadmap and Will Chen's posts before this framing is repeated in a sales context. If Mike ships a delegation layer, the competitive table in the manifesto §7 requires revision.
Source. Adapted from Craig Miller's LinkedIn essay "MIKE — The Demo Nobody's Watching Carefully Enough" — original at <https://www.linkedin.com/pulse/mike-demo-nobodys-watching-carefully-enough-craig-miller-azbye>. Body not available for direct quotation; this post is a faithful synthesis of the title's framing, the AGORA council's salience analysis, and the manifesto §7 competitive positioning.