EyeFly Infrastructure: TaskOS

TaskOS

EyeFly Digital · Akash · Leadership · v1 · 2026-05-06
01

How It Works

TaskOS is how EyeFly captures, organizes, and executes work. Every session, meeting, and Slack message feeds into a single Postgres brain — giving Akash and the team one place to see what's happening and what's next.

02

Layer 1 — Native Apps

LAYER 1 — NATIVE APPS
Where decisions, work, and meetings originate
Claude
  • ~80% of Akash's workday happens here
  • Decisions, builds & next-actions captured live
  • Every session ends with a structured handoff
Slack
  • Team DMs and channels hold inbound action items
  • Daily shutdown sweeps these for tasks
  • Support requests and escalations originate here
Granola
  • Records and transcribes every meeting
  • Speaker-labeled transcripts imported to brain
  • Action items auto-extracted on import
26
GCal
  • Calendar pre-fetched each morning at 7:40 AM
  • Meetings and blocks surfaced in daily brief
  • Reflects EyeFly's themed-day work rhythm
03

Layer 2 — Connectors

LAYER 2 — CONNECTORS
Automated pipelines — always running
Scheduled Tasks
  • Sweeps Slack DMs for action items 3× daily
  • Auto-saves stale sessions nightly at 11:30 PM
  • Runs TaskOS hygiene (dedup, archive) at 2 AM
Slack MCP
  • Reads team DMs and channels for action items
  • Posts morning digest to Akash at 7:46 AM
  • Sends P0/P1 task alerts direct to Akash
26
GCal MCP
  • Pre-fetches calendar events each morning
  • Powers 'what's on my plate today' briefings
  • Checks for scheduling conflicts before booking
Capture runs unattended. Session handoffs are the primary recall layer.
04

Layer 3 — Data Storage

LAYER 3 — DATA STORAGE
Everything persisted in Postgres
T&
Tasks & Projects
  • Every task attached to a project, every project to a goal
  • No orphan tasks — hierarchy enforced on write
  • Single source of truth: Postgres, not Notion
G(
Goals (OKRs)
  • Top-level outcomes everything maps to
  • Progress calculated from project completion
  • Reviewed and updated every Monday
SH
Session Handoffs
  • End-of-session state written by bye skill
  • Semantically searchable: 'what did I work on?'
  • Primary recall layer before any other table
MN
Meeting Notes
  • Full transcript + vector embedding per meeting
  • Action items auto-extracted and linked as tasks
  • Speaker labels preserved for full context
05

Layer 4 — Skills

LAYER 4 — SKILLS
Claude intelligence layer — on-demand or scheduled
SH
Session Handoff (bye)
  • Captures full session state on wrap-up
  • Writes structured row to Postgres with embedding
  • Records what was done, decided, and what's next
DS
Daily Sync
  • Morning ritual: processes voice notes or transcript
  • Routes items to tasks, calendar, projects or offload
  • Starts every day with a clear one-thing focus
DS
Daily Shutdown
  • Sweeps all team Slack DMs and channels
  • Captures every inbound action item as a task
  • Runs at 9 AM, 1 PM, and 5 PM daily
TH
TaskOS Hygiene
  • Deduplicates redundant projects automatically
  • Archives zombie tasks and stale projects
  • Advances pipeline stages when tasks complete
E
Execute
  • Reads recent handoffs and fans out parallel work
  • Applies 80/20: ships the 20% that moves the metric
  • Builds a dependency DAG before starting
06

Layer 5 — Frontend

LAYER 5 — FRONTEND
What the team sees and uses
AI
Akash Inbox
  • Morning brief: P0/P1 + today's one thing
  • Overdue items and escalation watch
  • Generated 7:45 AM M–F, read in Claude Chat
AP
Active Projects
  • All in-flight projects with % complete
  • Grouped by goal, shows blockers and next actions
  • Rebuilt from live Postgres data on every open
Slack Digest
  • Daily summary: ONE THING + calendar + projects
  • Posted to Akash's Slack DM at 7:46 AM
  • Replaces manual morning check-in
07

Vision

VISION — COMING NEXT
Autonomous Task Management

Activity in GHL & Slack automatically moves tasks to the right stage and creates new ones. Everyone knows what to do — without being told. Leadership freed from daily standup.