I've looked for a while, but I've never found a design like what I imagine: a single-tube (valve) regen CW transceiver (no cheating with a tube/valve that has two or more triodes/tetrodes/pentodes in a single envelope).
The way I imagine this working is that you'd receive like any other regen receiver used for CW ‐ oscillation of the regen detector provides a BFO to make the CW signal audible ‐ but when the key closes, the coupling between tickler and tuning coil is increased to a level that oscillates independent of received signal. Regen receivers are well known for causing "interference" on the tuned frequency if the regeneration is increased too far (oscillation couples back to the antenna, if there's no RF amplifier stage to prevent this).
A single tube/valve regen usually produces an audio output signal audible only with a very sensitive high impedance earphone, but that signal can be fed to any common audio amplifier to provide speaker output.
I'm interested in this in part because it's the simplest possible transceiver -- a dozen or so components (not counting a modern digital frequency display, necessary to ensure it's in-band and to tune a specific frequency), capable of running on battery power, potentially small enough to carry in a light backpack (including A and B battery). I know it won't be the best sort of circuit, which I presume is why I haven't found such a design (most regen designs put effort into avoiding emission).
@tomnexus found a listing for such a one-tube transceiver, with AM from a carbon microphone, but per the article linked, "It’s doubtful whether this simple circuit would meet the current FCC spectral purity requirements for use on the ham bands. After all, even while receiving, the regenerative receiver is radiating." Now, I know it's possible to operate a regen at a level of regeneration that doesn't radiate enough to cause interference, and with a dual tickler still get enough oscillation to transmit at low power, but there's still the question of chirp, drift, and harmonics to be concerned with.
Harmonics can be addressed by a low pass filter, either directly in the oscillator (on the transmit tickler loop?) or as part of the antenna coupling, but chirp and drift are harder to address.
Is this even possible, in terms of meeting FCC standards for spectral purity, etc., or in terms of staying on the same frequency for transmit that it's tuned to receive?