Sam Altman’s GPT-5 AMA: Bugs, Comebacks, and the Chart That Lied
OpenAI’s GPT-5 launch was supposed to be a triumphant leap forward—but instead, it stumbled out of the gate. In a candid AMA, Sam Altman addressed the rocky rollout, fan backlash over GPT-4o’s disappearance, and even a now-infamous “chart crime” that set the internet ablaze.
Here’s the breakdown of what went wrong—and what’s coming next.
1. The “Bumpy” GPT-5 Rollout
The biggest issue? A faulty autoswitcher—the system that decides whether users get the lightweight, fast version of GPT-5 or the full, deep-thinking mode.
When the router malfunctioned, responses felt… well, dumber than expected. Altman admitted the glitch made GPT-5 seem worse than it actually is and promised quick fixes, including:
✅ Smarter switching between model variants
✅ Better visibility so users know which mode they’re in
✅ A manual “thinking mode” button for those who want full power
“We’ll get this smoothed out,” Altman assured users.
2. GPT-4o: The Fan Favorite May Return
Reddit and X (formerly Twitter) lit up with complaints—some even joked it felt like “losing a pet.”
Altman heard the outcry and hinted at a possible comeback:
🔹 OpenAI is exploring ways to bring GPT-4o back for Plus subscribers
🔹 They’re weighing trade-offs but want to keep long-time fans happy
🔹 As a consolation, ChatGPT Plus rate limits were doubled, giving users more room to test GPT-5
3. The Infamous “Chart Crime
During the GPT-5 demo, eagle-eyed viewers spotted a glaring error: a bar chart where a lower percentage had a taller bar than higher numbers.
Altman owned the mistake, calling it a “mega chart screwup” caused by last-minute prep fatigue. The chart was quietly fixed in the official blog post—but not before it became immortalized in memes.
4. Altman’s Closing Promise
Despite the hiccups, Altman ended on an optimistic note:
“We’ll keep working to get things stable—and keep listening to feedback.”
For OpenAI, the GPT-5 saga is far from over. But if history is any indicator, they’ll likely bounce back stronger than ever.
Final Thoughts
This AMA was a rare peek behind the curtain at OpenAI’s growing pains. Bugs happen, memes spread, but the real test is how quickly they adapt.
What do you think? Should GPT-4o make a full comeback? And did you spot the chart blunder? Let us know in the comments!
