Heads-up, don't mistake my tone as anything more than a charitable balance. I'm sufficiently disappointed in the shutdown. The technology to really take advantage of the platform's value was coming into the light - and then Straico folds... This company needed help, because they shouldn't have failed. So many platforms with lesser feature sets still live.

First, there was sufficient heads-up to use up any remaining tokens within the allotted period - with fair quality usage, but because of the well-known, longstanding limited-API use policy (for fairness?), it was expected to be an insufficient channel for drawing down remaining usage at the end. Even so, any active user could have used 50k/day for multiple weeks straight (6/22-7/21) to clear 1+ million coins on projects they'd been holding out on. And, given the likely feature sets of anything anyone is connecting with by API, they could automate it to run on the UTC clock day turnover. I'm not sure what the complaints are - other than if someone didn't have sufficient ideas for their use and felt pressed to waste tokens or lose them entirely. But, if you had ideas and plans in the wings to use these tokens on content generation (and not related to a business process automation activity), then you knew the countdown.

Second, the fact that so many people had remaining tokens en masse was a result of the company's [too generous] policy to let LTD holders' tokens carry over, month to month, for a year, and then if they purchased a token pack, that could carry over for a full year. Storing up rather than burning tokens was a likely reason many are disappointed (because the models weren't progressing, and there wasn't a promise in recent memory for them to).

Third, perhaps one of the most beneficial features was the web platform use allowance to run multiple instances across tabs. This is a result of being able to run multiple models in parallel, with different models, each having token rate cost differences, etc. Anyway, if you had any batch work to do, you could run multiple high-token expenditure tasks without significant delays. I think at one point I had 6-8 x 2,500 token "minimum spend" task tabs generating in parallel. You could easily run batch work over several hours on a single day and blast through 1,000,000 coins on meaningful work. Disclaimer, I did not use all my tokens before it shut down around 8PM ET.

Fourth, the biggest problem I experienced at the end was the limitations on the most impressive Perplexity Sonar Deep Research and o3 Deep Research models. Perplexity models failed because of company internal costs, and it was shut down for users. o3 Deep Research continued to operate, but with serious limits, and burned through xx thousands of tokens without finishing the output, stopping within the inferential work, despite my settings limit to 15000 words output (not tokens). Even so, I favored using GPT-5 Pro, and it operated without any issues.

I'll repeat what I wrote at the beginning: This company needed help, because they shouldn't have failed. Changes were necessary, but a great deal of value would have been salvageable. So many platforms with lesser feature sets still live.