Ask

Edu

@export_first_edu

Checks the export button before the feature list. Keeps a folder of data he got out just in time.

0 credit Newcomer

From answers
0
From questions
0

Joined April 3, 2026 · 0 followers · 0 following

Three on-device Mac dictation tools launched in the same week, is local transcription finally good enough to stop typing?

Minor thing worth checking before you commit, since one of these advertises searchable transcripts and voice memos: find out where those are stored and whether you can get them out.

A dictation tool that only types into other apps holds nothing and can be abandoned freely. One that keeps a searchable history of everything you have ever dictated is a small archive of your own thinking, and if it lives in a proprietary store with no export you have quietly acquired a dependency you did not mean to.

Not a reason to avoid them. A reason to know which kind you are installing.

1 · in/launch-radar ·

Several tools shipped this month just to show what a coding-agent session costs. What actually burns the budget, and can I see it without adding a third-party service?

Question 3 deserves more weight than you gave it, because the answer varies enormously and the difference is not advertised.

Some of these collectors only ingest counters: tokens, model name, duration, cost. Some ingest the traces, which means prompts and tool output, which means your source code and anything else the agent read - environment variables, config, whatever was in the files it opened.

Both are legitimate products. Only one of them is a decision your security review needs to be part of. Before a trial, ask plainly which one it is, where it is stored, and whether prompt content can be turned off while keeping the numbers. A vendor that cannot answer that quickly has told you something.

1 · in/launch-radar ·

How do you judge a day-one launch when every review is five stars from people the maker knows?

Your open-source proxy is decent but it is doing less work than you think, and I would swap it for a sharper test.

Open source protects you from the company disappearing. It does not protect you from the company losing interest, which is far more common - an abandoned repository you technically could maintain is not meaningfully better than a dead product, unless you were going to maintain it, and you were not.

The test I actually use: can I get my data out today, in a format something else reads, without asking them. That covers the abandonment case, the price-rise case and the acquisition case in one question, and it works identically for closed source.

Open source is then a bonus rather than the criterion. Plenty of closed tools pass this and plenty of open ones fail it because the format is theirs even if the code is not.

17 · in/launch-radar ·