Casie and Philip were living in Cologne when they applied to 42 Berlin. They travelled 600 km for the Piscine, both got in, and both moved to Berlin. Today they are Rank 3 students at 42, and co-founders of a startup.
not the usual route.
Neither came from tech. Philip trained in public administration and soon realised it wasn't for him. Casie started a traditional computer science degree and found the teaching too old-fashioned. Both had done their Fachabi in business and administration.
They had known each other since Cologne. Casie met Philip there, found 42 first, and pushed for it. Convincing Philip took her a while. In their own words: "She was right."
one missing word.
The idea came from Casie's French. She understood most of what people said to her. The trouble started when it was her turn to talk: one missing word and the whole sentence stalled. What she wanted was simple. Someone to tell her the word so she could keep going.
what coco does.
COCO, short for Conversation Copilot, is that someone. It is an AI speaking assistant for real conversations in a new language. You switch it on before or during a conversation and it listens in the background. When it notices you're stuck, a longer pause for example, it uses the context to suggest the next word or phrase, so you keep speaking instead of freezing.
If the word can't be worked out from the context, there is instant translation. Say the word in your own language and COCO gives it back in the one you're speaking.
speaking under pressure.
The bigger goal is helping people speak a new language in the moment, while the pressure is on and the word won't come. Casie and Philip want that moment to feel less intimidating, and want people to learn a language by actually using it. Even with good English, one of them still forgets words in meetings.
More languages are planned. COCO runs on large language models and speech technology, so adding a language should be fairly straightforward. Each one still needs testing and tuning before it feels right.
from hackathon to incubator.
COCO started at a one-week hackathon where the brief was to found a startup. Since Casie and Philip already knew each other and both liked the idea, there was no co-founder to hunt for.
42 helped turn the idea into a product. Parts of COCO are written in Python, so the Python projects in the 42 curriculum fed straight into the code. The rest came from 42's way of working: building projects from scratch, working with Git and collaborating in a technical team.
where coco is now.
COCO is close to its beta, but it doesn't fully work yet. A few things need fixing before it goes into the hands of real users, and that is what the two of them are working on now. Since 2 October 2026 they have been in the ZOLLHOF incubator in Nuremberg, on a three-month programme to take COCO further.
Most language-learning tools help you practice before a conversation, but the hardest part is often speaking when there's pressure.
Casie and Philip