Skip to content

Releases: pedroromeroluna/ai-first-os

v1.3.0

Choose a tag to compare

@pedroromeroluna pedroromeroluna released this 27 Aug 03:11

Changed

  • The nine pack skills that produce a deliverable resolve their destination without naming the
    system. With no product node loaded, the file goes to the current folder and the skill says the
    path it wrote and nothing else — no new-workspace, no new-product, no session scan, and no
    mention of AI First OS. Someone who installed a single skill from a link no longer hears about a
    system they do not have.
  • prd writes its eight strategy files into product-context/ when no product node is loaded,
    instead of leaving vision.md, problem.md, users.md and the rest loose in whatever folder the
    session started in, where they could overwrite files that were already there.
  • synthetic-users names its deliverable <YYYY-MM-DD>-script-diagnosis.md; it was the only one of
    the nine naming its file in Spanish.
  • The pack offers its sibling skills once per session, not at the close of every deliverable.

Added

  • new-memory, a command that creates a memory-node type the first time it is used — the folder
    plus its five lines in tree.md — and adds a node to it. Two questions, what kind of thing this
    is and what this particular one is and for whom, and it writes the head of the node. Until now a
    brain could only grow the node types the system already knew.
  • The spec craft records the commit of main the spec was written against, so the agent that
    implements it can see whether its terrain moved, and states that the spec is committed to main
    as soon as it is written — which is what reserves its number when two sessions write at once.
  • The contract states that a branch declares itself finished in a file, never in a chat, and that
    whoever merges removes the worktree and deletes the branch.

v1.2.0

Choose a tag to compare

@pedroromeroluna pedroromeroluna released this 23 Aug 01:59

Added

  • rename-heads, a command run only at the operator's request: moves a brain born before the single
    node shape to it — every node becomes a folder with its README.md inside, every initiative gets
    its own folder, and each product's dated documents move from context/ to research/ — rewriting
    the head paths wherever the system reads them as routing, leaving URLs untouched, and listing what
    it does not touch for review.

Changed

  • One shape for every node: a folder with the name of the thing and its head README.md inside, at
    every height — workspace, product, initiative. The root does not change: its folder is the brain
    and its head is operator.md. A brain already installed keeps working with no change: the shape
    of the head is read from its own tree, never assumed, and the previous one stays valid
    indefinitely.
  • A product node keeps context/ for one meaning only —its strategic layer— and every dated
    document about it (the pack's briefs, analyses, inventories) now lives in research/.

v1.1.0

Choose a tag to compare

@pedroromeroluna pedroromeroluna released this 22 Aug 15:26

Added

  • rename-workspaces, a command run only at the operator's request: adopts the current name of the
    workspaces folder on a brain born before this rename, rewriting the prefix everywhere the system
    reads it as structure or routing and listing what it leaves untouched for review.

Changed

  • The isolation boundary is named workspace everywhere the operator reads it — the interview,
    the manual, the README — replacing "organization". A brain already installed keeps working with
    no change: the folder name is read from its own tree, never assumed, and the previous name stays
    valid indefinitely. The command that creates one is now new-workspace; --org keeps working
    everywhere, with --workspace as its exact synonym.

v1.0.0

Choose a tag to compare

@pedroromeroluna pedroromeroluna released this 22 Aug 02:45

Added

  • The brain: a plain-text folder versioned with git, built by an interview that writes who you are,
    how you write, your root resolver and the tree of paths a scan walks.
  • The session contract: every session opens with a scan of what waits on you, what is in progress
    and what is queued, plus the checks, and with the startup read of the files that give it context.
  • Ten skills of the system, reached through the resolver and never activated on their own:
    bootstrap, new-org, new-product, sweep, check-resolvable, capture, close-session,
    mount-repo, grill and new-spec.
  • Organizations as the isolation boundary, with their own identity, resolver, backlog, decisions,
    learnings, initiatives and products.
  • A repo mounted onto an initiative — an existing remote, or a new repository born from the
    scaffold — with its row in the environment table.
  • Delegation to agents on a branch: two implementers picked by the state of the spec, plus a reader
    that never writes.
  • The pack of product skills, published on its own in the Agent Skills format and offered at the
    end of the install, or installed later from the backlog.
  • Install by symlink, idempotent, with git and python3 checked before anything is written; updating
    the repository refreshes every brain that points at it.