@LLMJunky@michael_kove I blew through $100 in 5 “design” prompts from Vorflux delegating to my GPT models. It’s rough out here and I agree with the above statement.
@sama I know you’ve probably seen this enough, but I’d really like this available in my desktop app to delegate out tasks and orchestrate my codex chats.
Note how he says “modern public frontier models” - roon, I hope that GPT-6 absolutely fries and releases relatively soon pending no one says “WORLD ENDING MODEL IS COMING SOON”. We don’t need the USG stifling our fun ;)
the era of the chinese labs being far behind is over, Kimi is at least on par with the modern public frontier models. people have to think differently now without any competitive margin built in
Hi Jason, I’m Mike and I’ve been in IT for 15+ years working on support, networking, systems and infrastructure between the private and public sectors.
I use codex to help me automate the mundane work that generally speaking, people really don’t want to deal with. It’s been my exo-suit and now I want to build a system that cuts through ambiguity and drives value by creating further workflow optimization and refinement for the SMB logistics slice of the economy.
Am I the only one who thinks this guy is off his rocker? No shot.
Has anyone mentioned the models being held behind closed doors to this guy?
Tinfoil hat time: Mythos has been locked away from the GA for 5 months.. GPT-6, etc. My assumption is that we’ve hit the flywheel so hard that the USG and AI leaders have had to keep the recursive self trained stuff on the DL and continue providing dumbed down versions for us normies.
Today, we are introducing Inkling.
Inkling reasons efficiently across text, image, and audio modalities. We are making the full weights available.
https://t.co/Ghebq5mG30
Available today for fine-tuning on Tinker. Play with it in the Inkling Playground. 🧵
🤯 Genuinely surprised to see that when having 5.6 setup @vorfluxai custom harness, it selects Anthropic models in reasonable areas. Kudos to intelligence when it recognizes game.
@trikcode Personally, I built a "memory-spine" across all my agent harnesses. There is cross validation across my repos and implements into a wiki based structure that is run 4x times a day for freshness. Each of my harnesses feed into this to ensure there's no lost context at any point.
@OpenAI - I really enjoy starting off a strong research task or something I feel requires some heavy/deep thought and then opening the same context window and switching to 5.5 Live to go over that context on the phone to jog the brain a bit further with voice. Incredible!
Truly next level!