It appears that there has been a noticeable shift in focus towards API development at the expense of advancing ChatGPT-related offerings for customers. Reflecting on just a year ago, updates were more frequent—even weekly—and it’s likely that the budget was even smaller at that time. The pace of innovation has slowed considerably, which some might attribute to overly cautious risk management. Indeed, our progress has traditionally been propelled by a trial and error approach, yet today’s models still show significant limitations, suggesting that perhaps this cautious approach is now more suited to research teams. For instance, in my personal projects, like scripting a goat simulator role-playing game, the advancements seem minimal.
It’s worth noting that the most recent update was around mid-February. I would humorously suggest skipping directly to GPT-5 if it meant faster progress. The current security measures, while intended to safeguard, can often be circumvented by those determined enough, leading us to question the trade-off between security and performance.
Additionally, if security is a primary concern, it might be worthwhile to revisit insights from previous team members who were adept in risk assessments. Unfortunately, some of these valuable contributors, like the gentleman who was notably bald (I share this trait), are no longer with the company, which is a significant loss.
On a lighter note, the situation somewhat reminds me of Sam Bankman-Fried’s tenure in the Bahamas; a period marked by a lack of significant product evolution, allegedly due to internal distractions. It’s crucial to remember the trust that customers place in your company, hoping that you work ‘with’ us rather than merely ‘for’ us, as was the case previously.
Sending a warm and thoughtful greeting to everyone involved or not with the project.