
Why we built Smol
One recording, one audio file, six thumbnails. The part missing from the ZIP was the conversation about the edit.
One podcast handoff we used to build Smol arrived as a ZIP. Inside were a MOV, a WAV and six thumbnails. You can send an editor every byte in that ZIP and still leave them wondering which cut to use.
We already had a chat app and a document drive. Putting a conversation inside each project folder gave the editor somewhere to find the decision as well as the recording.
Eight files made the problem concrete
The six thumbnails could finish unpacking while the recording was still coming in. If you’re picking cover art, you can already compare the candidates. An import screen that says only “working” makes you wait for a file you don’t need.
So Smol lets you open, rename or move each thumbnail as soon as it’s ready. By the time the recording is ready, the cover can already be picked.
Put the discussion beside the files
An editor needs to know who’s reviewing the cut. That question belongs in the folder conversation, where the whole team can answer. A correction to a particular passage belongs on the version the reviewer actually saw.
An example episode in a local build. The next person can read the conversation and open the cut from the same screen.
That distinction matters when a second cut arrives. The reviewer’s old comment still opens the first cut at the passage they meant. Nobody has to guess which edit they were looking at. An owner and a status on the work tell the next person where to pick it up.
Find the quote, then check the source
Say you need a line from the interview for the episode intro. A filename can tell you which recording to open, but it won’t tell you where the guest said it. Smol transcribes supported recordings and indexes the transcript. Ask this file can point you back to the source in the version you selected.
We checked that path with a 102-second synthetic WAV. The transcript was ready and indexed by the end of the test. We didn’t time how long that took. If transcription fails, the editor can still open the original.
Keep the progress when the connection drops
Big recordings raised another question. If an upload is almost finished when the tab reloads, how much does the editor have to send again?
We reloaded Chrome after 18 of 20 parts of a 200 MiB file had arrived. We picked the original file again. Chrome sent the last two parts, 20 MiB in all, and the downloaded file matched the original byte for byte. The check and final commit took about three seconds. Our measurements start after the reload, so they leave out anything already in flight when the tab closed. Browser measurements · How Smol resumes an upload
Straight upload speed gave us a less flattering result. In our first two 100 MiB trials on one Mac, both Drive and Dropbox finished ahead of Smol. The chart shows both runs for each service, measured from the first payload to the final commit. The accounts differed and we didn’t control background traffic.
We later let the browser send up to four parts at once. That cut elapsed time in our local tests across four kinds of uploads. A separate 100 MiB production upload used three parts at once and finished in 19.4 seconds. We haven’t repeated the three-service comparison with that uploader. See the upload timings
Start with one project
If you’re trying Smol with a teammate, pick an edit that’s already in progress. Put the current cut in a folder and leave the editor the one thing they need to do next. When the next version comes back, see whether you can follow the decision without asking anyone to find the old thread. Open Smol · Get the Mac beta
Choose the folder’s sharing policy before you bring in private work. Personal and new team workspaces start private. A team can allow unlisted links under folders that permit them. Smol is in alpha, operators may read stored data, and data may be lost. Keep a backup of your originals.

