A podcast production workflow that runs itself.
Note 2 Self was built as a working production system, not just a pilot. Its independent podcast producer faced a bottleneck that wasn't making the show — it was everything that happens after recording.
The producer runs the show alone. Each episode generates a second job: pulling social clips, burning in captions, preparing assets for distribution. That work is repetitive, time-consuming, and sits directly between finishing an episode and publishing it. It also scales badly — the more episodes you make, the more of your week it takes.
We built an agent the producer operated through Telegram. She sent it what she needed; it handled clip selection and caption processing against her existing files. There was no new software to learn and no platform to migrate into.
The distinction is between making an episode and getting the material around it ready to use. A useful assistant had to work on the actual recordings and captions, not just suggest what a producer might do next. The engagement focused on that repeated production work while keeping the existing files and a familiar messaging interface.
| Project context | Public detail |
|---|---|
| Client profile | Note 2 Self — independent podcast producer |
| First conversation | May 15, 2026 |
| Production environment live | May 21, 2026 |
| Fully connected | By June 1, 2026 |
| Measured feedback | June 25, 2026 |
| Interface | Telegram — no new application to learn |
| Deployment | Dedicated production environment |
After recording came a sequence of smaller jobs: finding moments that could stand alone as clips, preparing captions, and assembling material for distribution. For a solo producer, those jobs draw on the same time and attention needed to make the next episode.
Working with isolated tools also creates handoffs. A clip may be selected in one place, captioned in another, and moved again before it is ready to use. Even when a tool can perform an individual task, the producer still has to supply the files and carry the context between steps.
The initial scope was therefore practical: work against the producer's existing material, handle clip selection and caption processing, and make the system accessible through a message. It was a production-workflow engagement rather than a change to the show's creative direction.
Telegram is the operating interface. The producer can ask for work in a message rather than translate the request into a new dashboard. This keeps the point of interaction familiar while the dedicated environment does the processing behind it.
The system identifies social clips from episode material. That addresses the repeated search through recordings for usable segments, rather than merely generating a list of promotional ideas. Clip selection is attached to the existing episode files, so the request is grounded in material the producer has already made.
Captions are burned into the clips as part of the production pipeline. Selection and caption processing are connected steps rather than two unrelated jobs the producer has to set up from scratch. The output is prepared media, not just instructions for another editing session.
Workspace access connects the system to the files the producer already uses. She does not have to migrate the show's material into a new application to ask for processing. Connecting that access was a distinct delivery milestone after the initial production environment went live.
A dedicated production environment provides the place where the workflow runs, separate from the Telegram conversation. The delivery also established a deployment path and update pipeline, so the engagement included a way to maintain the system rather than only a working first demonstration.
Discovery began on May 15, 2026. The dedicated production environment was live on May 21, six days later. That was the first deployment milestone; it should not be confused with the later point at which all the workspace and update connections were in place.
By June 1, workspace access, the deployment path, and the update pipeline were connected. That seventeen-day span covers the move from the first conversation to a system connected to the producer's working environment, not just an isolated process running on a server.
The June 25 feedback conversation came forty-one days after discovery. It supplied a different kind of evidence from a successful deployment: the producer's account of whether the system was useful in her production work. Infrastructure records subsequently documented the environment as active in August 2026.
Timeline reconstructed from infrastructure records and recorded conversations through August 2026. Deployment dates are distinct from client-reported outcomes; workload savings remain an approximate estimate.
“It's incredibly helpful. I wouldn't be able to do the velocity of what I'm doing without it.”
“The majority of it is, like, perfect.”
She estimated the system reduced her weekly production workload by roughly a third — time returned to making the show rather than packaging it. That figure is her own estimate, offered as approximate.
Her feedback was about production capacity, not a benchmark score. The assistant was useful because it took on part of the recurring work around each episode, allowing her to sustain a pace she said she could not otherwise manage.
The approximate saving should stay approximate. It was reported by the producer, not established through an independent time study, and she asked to confirm the precise hours before publication. The record supports a meaningful reduction in workload without turning that estimate into a guaranteed result for another show.
The system doesn't make the show. It handles the part of the work that was standing between finishing an episode and publishing it.