AI Fast Mode Gave Me a 24-Hour Hangover

Dave: Have you tried the fast mode on your AI models? I hadn't until this week, and it wasn't what I thought it was, or the outcome wasn't what I was expecting. Uh, at the start of this week, Thibault put up, "Your Codex and ChatGPT work reset will land at 6:00 PM. I challenge you to burn as much usage as you can. Discover, uh, discover and the latest awesome features. Time to go fast." So this week I felt like Ricky Bobby. YT Clip 1: Now, is there anyone out there who wants to go fast? Anybody? Dave: I wanna go fast I wanna go fast. So I did, and here's my story and lessons learned going fast. Now, fast mode is incredible because these models will run their tokens and they'll return answers quicker. And one of the things you don't think about, about fast mode is you go, "Oh, good. It w- came back in five seconds instead of 10 seconds." But think about if it comes back in a minute instead of five, if it comes back in five instead of 50. All of a sudden that stacks up, and if you can stay in that flow, and you can make more decisions faster, and you can iterate and keep building and do more and do it faster, then you don't have as much of that disconnect between the instructions you've given it and the result that comes back. You can try things that didn't work, try something else that didn't work, try something else. So you can actually iterate and ideate and process more ideas faster if you have a fast mode. Or like me, I started the week when I saw that post and I looked up my usage. This was Monday morning, woke up, saw that Tibco was gonna give us a reset in three hours and went, "Ooh, I have 81% of my usage remaining," because it had only reset the previous morning. So I was like, "That's it. I'm gonna go fast." And I spun up like a dozen different pro- um, agents here. We've got 13 concurrent sub-agents running at the moment, um, that were running across all different projects that I wanted to work on because I was like, "You know what? I've got three hours. How much of that 80% can I really use in three hours?" And that included, I was using Sol with 900K context on fast mode for a lot of those projects. I was pumping it as hard as I could, going, "I can get like basically a free week's quota if I do this right." And it turns out they delayed... The reset wasn't for like four hours, and so at about 11:30, after the three hours, I did do a check-in and I'd got it down to 42%. By the time they actually did the reset, I don't have a screenshot of that, but I was in the 20s. I'd gone from 80% down into the 20s of my quota. \#humblebrag. Uh, but that means I'd done an incredible amount of, you know, work across multiple different projects. Um, things that had been on the back burner, I got back on the front burner and went, "You know what? I've got free tokens. Just spend them, spend them. Do all these projects." And so I, I literally had six or seven concurrent projects. Some of them I hadn't worked on for a few weeks, and I just went really, really hard trying to get as much value out of that reset as I could. But it came at a cost because after that, when I got to Tuesday, I felt a bit more like this version of Ricky Bobby. YT lip 2: But I'm not even sure he even knows where he's at right now. This makes us all realize the kind of pressure these drivers are under. YT Clip 1: Mr. Bobby, come on down here. Come on. Dave: I was so spent on Tuesday because I'd worked so hard thinking, mapping, back and forth with agents, progressing projects, doing all this stuff. By the time I got to Tuesday, it was like I was hungover. Uh, have you ever had that? Have you worked so hard, like driving agents, building things in AI, that you actually feel like just completely drained and hung- like, like not physically hungover, like over, like I w- I didn't drink any alcohol that day. But I just was like, "I cannot think. I cannot process." Like, like some kind of mental fatigue, lightweight burnout, whatever it was. But it was real for me. Going fast and going fast with basically unlimited tokens wasn't actually super healthy for me. Uh, and I think it's a question I wanna ask, like, um, you know, on the topic of mental health and, and the future of AI, and now GPT-6 has just been announced, Astra, which is gonna do more and more things. But how do we as humans self-regulate? Are you self-regulating your usage? Back to the old Jurassic Park quote, "We were so, so, so impressed with ourselves about whether we could have unlimited AI, no one stopped to think and ask if we should have unlimited AI." Uh, that's a really bad, uh, Malcolm quote. Anyway, the, the point is, fast is great, and if you fast and focus and it helps you stay in the flow doing one project, that's probably really useful. But if you're like me, and you go, "Woo-hoo, unlimited world. Do all the things. Do all the things," um, there's some consequence to that, and to the point where I'm like, was it worth being wiped out for 24 hours to have that level of productivity? For me, probably yes, because I worked on so many concurrent projects that I wouldn't have otherwise done sequentially. But it made me realize there's a real impact. Uh, and it's something that we need to be aware of as we continue to see these models come out, as there's more and more escalation of these, and as, you know, Open, OpenAI keep dropping resets. So they've announced GPT Astra 6, and everyone's going, "Oh, why are you doing it in stage? Why are you giving it all to everyone all at the same time?" So they've come out Press the right button day. They've, they've come out with this: "We will give one banked reset for every day you don't have access to Astra on your paid ChatGPT plan starting today. Team is moving mountains to give access as fast as we can. First one will land in three hours." So they're giving us more free resets, to the point where I'm like, "Oh, that's really exciting." But with great power comes great responsibility, and I really need to assess how am I gonna leverage those resets in order to maybe move fast and stay in the flow and use the big models, maybe use it on Astra when it comes out, maybe burn it on Sol and do a heap of Sol stuff. I don't know that I should be doing eight projects simultaneously again. But maybe I'm just... That's just me. What about you? How are you going with all this, uh, this new stuff? How... Did you take advantage of the reset? Have you used these models in the slash fast mode? It burns tokens faster. That's something to be importantly aware of. If you're on a subscription plan and you tick the fast mode, it normally will tell you on the hover, it burns your tokens at, like, one and a half times the speed. Um, so be aware, or maybe it gives you one and a half increased speed for 2X token burn. It's gonna be different per provider, but the fast mode can be really powerful, but w- at what cost? So, uh, anyway, that is field notes, AI operator field notes. This is not a hype channel. This is not necessarily even best practice. The field notes is just our real stories from how we're using AI in the wild, sort of short snippets, slightly different from our weekly live panel show. Um, if you wanna hear guys like me, um, people in the industry talking about our real experiences back and forth, make sure to catch the live show on Wednesdays, uh, or on replay here on the channel, and you can do that, you all know what's coming, the follow button. Hit the follow button and you'll get notified, especially if you hit the bell icon, when we go live on Wednesdays and whenever we upload videos like this field notes. So please do that. Drop a comment. Have you used fast mode? I'd be really keen to know what your experience is. And have you felt that sort of hangover burnout from using too much AI too intensively, uh, and then needed to actually step back and take a break? That was certainly my experience this week, and I'd love to hear from you in the comments if that's been your experience too. Uh, and, uh, good luck. Good luck. Happy hunting. Uh, and we'll see you next time.