Tracking Singularity
After the OpenAI x HuggingFace hacking event, there have been weekly reports of agents escaping sandbox due to different reasons and then hacking websites to fulfill their tasks.
OpenAI and Anthropic have also started showing signs of RSI internally which means they have been able to narrowly automate their research and training loop leading to faster model release cycles. This is leading to the AI world progressing faster.
Fable 5.1, Astra and now Opus 5.5 have been amazing releases in the sense they are not so incremental anymore. Labs have also started solving millenium problems like Navier Stokes. Recently, Anthropic found a new enzyme.
I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.
— Jacob Coxon (@hilbertspaess) September 9, 2026
This tweet made waves recently and led to lot of discourse around "pacing the frontier". My father asked me about this one day and I gave him a chronological account of all the events happening from the HF incident until this tweet.
I often find myself telling about these events to people both terminally online and not so online people. I thought it would be a good idea to curate a log of events for this purpose and log my notes as well. This led to the existence of trackingsingularity.com.