Model routing layer
Route work across coding, research, triage, summarization, and image lanes.
A personal agent infrastructure that runs my real life, rebuilt on Meta models
Suite Fleet is a working collection of agents on one small server. This hackathon build puts Meta models into the daily jobs the fleet already handles, then measures the results in public.
Built from a live workflow
One server.
Real daily work.
Measurable model choices.
01 / THE STORY
A fleet that already has a job
The fleet runs on one small server and helps with the work that fills an ordinary day: triaging mail across several accounts, scoring job leads, watching deals, and keeping study routines moving. It also supports a personal suite for mail, notes, files, and chat.
Agent seats coordinate through an authenticated room. A coding lane turns requests into software changes, while an independent reviewer seat checks the work before release. The owner stays in control: money, messages to other people, and submissions pass through hard personal approval gates.
02 / THE HACKATHON BUILD
Four pieces, one working system
The fleet and its workloads already exist. The ten days are for integration, measurement, and polish.
Meta model lanes
Put Meta models to work on real daily jobs inside the existing fleet.
Public results dashboard
Compare cost, latency, and accept or reject verdicts for each task type.
E ink day brief
Bring the day's most useful fleet output onto a quiet, glanceable display.
03 / ARCHITECTURE
Clear roles, deliberate handoffs
The owner directs the fleet. Specialist seats and model lanes do the work; the reviewer checks changes before they move forward.
Approval gates remain with the owner for spending, outbound messages, and submissions.
04 / VISUALS
A place for the real screens
These frames are reserved for public exports from the build.