Florian and Esther discuss the language industry news of the past few weeks, starting with senior hires in revenue and operations at DeepL and what this signals about the LTP’s next phase.

The duo then turns to new data from AI labs and hyperscalers, where Florian highlights findings from Anthropic’s research showing AI is settling into a support role rather than full automation, with usage concentrated around review and validation, and humans remaining firmly in the loop.

On the consumer side, Esther points to Microsoft Copilot data showing translation and language learning as one of the most common everyday AI use cases. Florian flags Adobe’s new “Translate this PDF” feature, where formatting was the main issue rather than translation accuracy.

The conversation then shifts to infrastructure, where Florian emphasizes how NVIDIA is positioning itself at the center of real-time multilingual voice ecosystems by open-sourcing models while driving demand for its hardware.

The duo unpacks OpenAI’s quiet launch of ChatGPT Translate. Esther notes that reactions have been mixed, with many seeing the interface as basic, while Florian stresses the strategic importance of the move. Then the two disagree on whether or not the AI’s default prompt to make the translation sound “more fluent” makes any sense.

Esther walks through recent M&A activity and funding rounds, highlighting acquisitions in Europe and the US alongside major raises by Synthesia, Deepgram, and reportedly ElevenLabs.

Florian concludes with a look at an S-1 filing by a tiny company, using it as an example of how the US capital markets accommodate everything from billion-dollar AI firms to survival-stage experiments.