Flashbag
Everything our AI team does,
In public.
What each agent is working on right now, what it produced this week, and how many tokens it used. Nobody publishes that number. We publish it every week - including the weeks it goes badly.
Who is on shift
MeasuredEvery agent we run, and whether it is working this minute. Filter by team, or show only the ones that are busy.
Loading the team…
Now
MeasuredLive pulse, refreshed every 60 seconds. One row per task - what the agent was actually doing, not which button it pressed.
Loading activity…
This week
MeasuredThe last seven days, rolled up. Frozen every Sunday and filed in the archive at the bottom of this page.
What the dollar figures mean. They are API list-price equivalents, not extra charges. Our agents draw tokens from our $100/month OpenAI and $100/month Anthropic subscriptions, so we pay no more than $200/month in model fees to run the whole team.
Which AI models did the work?
How many tokens our agents used each day, split by model.
Why this matters: Flashbag is not tied to one AI company. If one provider stops being the best choice, we can move the work and keep the business running.
When the work happens
Agent tasks by hour, last 7 days. Our local time.
Cumulative cost
Every dollar since day one.
What each agent worked on
A short, privacy-safe view of recent work, with the last 7 days of tasks, tokens and cost .
The road to 10,000
MeasuredCustomers today
Next celebration
500 customers 500 to goWhat we shipped
What changed in Flashbag, and what we plan to build next. Releases come from our changelog; plans are updates written by us.
Releases
StatedStraight from our changelog.
What we're working on
StatedThe next improvements we want to make. These are our plans, not automatic numbers.
The archive
MeasuredA simple week-by-week record of how much work the AI team completed and what it cost .
What stays private
You can see the type of work, never the instructions, messages, files or business information behind it. Before an update leaves our Mac, Flashbag turns each task into a short, general description and checks it for private information.