Teknium
A central figure at Nous Research, specializing in open-source model fine-tuning and dataset construction. He is recognized in the open-source AI community for consistently high output volume and quality, contributing substantially to projects such as the Hermes model series.
Timeline (last 90 days)
@Teknium found that Jev's compression tool score can be reproduced with one line of code (removing the tool call), and that the effect decays: the first round frees half the window, the next round frees a quarter, and then less.
@Teknium's data shows that comparing Jev compression with ordinary summarization: 115K tokens retained versus 55K, recall 76% versus 79%, and under the same budget Jev ranking ties with the ordinary recency method (77.8 versus 77.8).
In its announcement, Perplexity paid tribute to Teknium and the Nous Research team, praising their consistent product execution and taste.
Teknium is a core figure at Nous Research, with a long track record of output in areas such as open-source model fine-tuning and dataset construction.
All Facts (4)
@Teknium found that Jev's compression tool score can be reproduced with one line of code (removing the tool call), and that the effect decays: the first round frees half the window, the next round frees a quarter, and then less.
50%Unverified@Teknium's data shows that comparing Jev compression with ordinary summarization: 115K tokens retained versus 55K, recall 76% versus 79%, and under the same budget Jev ranking ties with the ordinary recency method (77.8 versus 77.8).
50%UnverifiedTeknium is a core figure at Nous Research, with a long track record of output in areas such as open-source model fine-tuning and dataset construction.
50%UnverifiedIn its announcement, Perplexity paid tribute to Teknium and the Nous Research team, praising their consistent product execution and taste.
50%