I’m ok with GPT-5, what I dislike about the whole thing is how OpenAI took the model use decision out of my hands… I’ve re-enabled legacy models, selected 4o, and it still switches back mid-chat.
So even if 5 is objectively better, my strongly adverse reaction to “we know better than you”-ism has resulted in me using Claude almost exclusively since 5 launched…
Granted I’ve also been in deep dev mode on 2 different software projects alongside my long-running hardware-based project, so that’s $$$ with Claude Code not chatting on claude.ai…
(also, hi! I’ve been in DEEP MONK MODE for a long-ass time, building multiple “escape pod” projects, any one of which will make me financially free… but why build 1 when I can disappear completely and build 3 instead?? )
‘Beloved’ is funny. And does seem very ‘on brand’ for how 4o was communicating sometimes.
Truth is, in terms of how I use ChatGPT there has not yet been any discernible difference between the two. I loved the previous version and I’m still loving this one.
I suspect that I give less room for its ‘personality’ to make much of a difference. I don’t tend to use most of its suggestions or necessarily follow down the paths it proposes. Still figuring out how to best interact with it.
This is what I wrote to two friends the other day. (Not completely related, because it was one of those ‘AI replacing people’ conversations. But it does express how I’m currently relating to AI.):
I think of AI, and relate to it, as a highly-responsive mental/organizational prosthetic.
I don’t ask my GPS (a navigational-AI prosthetic), “Where should I travel?” I choose where to travel and then allow my GPS to aid my route-setting. And it tends to be pretty excellent at that.
We’re the ones subconsciously driving this ‘AI will replace us’ narrative. It’s like (on some level) we almost want to be replaced or something.
Make choices. Use your imagination and capacities. And you are, by definition, not replaced.
That’s how I see it.
So, I think I’m still mostly relating to it as an extension of my ideas and my self.
A good movie that mirrors what’s going on with AI right now is The Social Dilemma. Tech grew so fast with smartphones and everyone was innovating. But it was like the bystander effect, everyone else assumed someone was going to keep it on the rails and prevent it from morphing into something harmful. We can see how much damage that’s done to people. I feel as if nothing was learned from that lesson or people generally aren’t caring that it could go that direction again.
Your take on AI is a balanced one, I just worry for those who don’t have that perspective . I read an article recently about a man who listened to chatgpt and replaced his salt in his diet with sodium bromide and slowly poisoned himself. That level of instant misinformation is terrifying.
When I read your post, what it seems like to me is a statement of your Vision, your Mission, your Calling.
Not necessarily precisely. And not in some straitjacketed, limiting sense.
But to me, I’m beginning to see that:
The perception of a problem is an invitation to your Purpose.
The fact that you see things like this is actually a cause for celebration. It means that you’re clarifying Problems that your mind is here to address and to help with.
Viewed from one vantage point, every career is a set of problems that we help to address.
I built https://qubictactoe.com/ using just Claude and Claude Code. 0% coded by me. My role there was more as “Software Architect” or something.
I usually use Claude Pro (claude.ai chat) to tell it an idea…we go back and forth and work out the idea and work out any kinks we can think of.
Rather than have Claude Pro do the actual build, I have it distill it all into a build plan saved as a markdown file.
I then fire up Claude Code on my server and give it that markdown file…it starts building right away, and asks any clarifying questions as they come up.
It’s brought a lot of the fun back into building things for me!
It’s set so when you play against the “AI,” it learns from losing moves when you win. Can’t beat it with the exact same strategy more than once usually. Mwahahahaha lol
Another game I made a couple days ago at work. Claude Code as per the usual, took total of 17 minutes. Told it I wanted a rip-off/clone of Galaga, first run took 10 minutes. last 7 was just tweaks
Wondering if anyone here runs local LLMs? Or experimented with them?
I’ve been testing a lot of local models. I just found out about pi which you can use as an agent for coding so I’m gonna see how far I can get with that.
My rig is pathetic (3GB vram) lol. Def don’t have the VRAM for more heavy tasks but it’s been fun running things local and doing stuff like updating resumes, journal prompting, and just being able to comfortably use my own data since it’s 100% private.
It’s honestly crazy to me some smartphones can even run local llms for tasks. Not mine though lol. But the way these larger enterprise models can be condensed down is impressive. Obviously it can’t compare to massive data centers and computing power but I’m genuinely amazed at what you can do even on CPU and RAM alone.
I want to. But my sole computer right now is a laptop from 2014 that originally shipped with Windows 8.1!
No real GPU in it. It won’t even be able to run the Slack desktop app anymore close to the end of the year since they’re upgrading their hardware requirements.
Something with lots of RAM and vRAM. I’m not married to any particular architecture. I worked at Apple so know my way around macOS, but I prefer Linux if not using Windows.
)
