Philip Nora

Running GPT, Claude & Grok agents daily and sharing what actually works. Real setups, real usage numbers.

[Philip Nora @PhilipNora7]

First-day review of Droid by @FactoryAI: mission mode, the CLI, how models behave inside it, and what I'd love to see improved. https://t.co/3a5Gz3cS0V

[Philip Nora @PhilipNora7]

Shoutout to these free models again. Made a cool audio visualizer while I was doing work and connects directly to my spotify

[Philip Nora @PhilipNora7]

I gotta say, I'm really impressed with the capabilities of these free smaller models in Hermes. Been using them for a few days and I'm very happy with them. I was using Luna max through my GPT plan and I was constantly hitting usage limit so I gave these guys a shot and for what I do with Hermes nowadays I don't think I'm going back to Luna.

[Philip Nora @PhilipNora7]

@ConnorTalksAI lucky mfs

[Philip Nora @PhilipNora7]

@levainbey It's fun to use for sure. Very low hallucination rate and great at catching and fixing bugs

[Philip Nora @PhilipNora7]

I’m really impressed with Gemini 4 Carbon

[Philip Nora @PhilipNora7]

For my people using Hermes. What are the best skills, plugins, and connectors you've used or are currently using?

[Philip Nora @PhilipNora7]

@ritu_twts @grok Ngl, right now G4 Carbon is better than opus 5.5

[Philip Nora @PhilipNora7]

@Da7_Tech Man, Fable 5.2 looks incredible

[Philip Nora @PhilipNora7]

@Da7_Tech Well maybe if us consumers weren’t hogging all the ram it could work better!

[Philip Nora @PhilipNora7]

@Da7_Tech Dude even better idea, make it a section on your benchmark website

[Philip Nora @PhilipNora7]

@soareso I’d be okay with that

[Philip Nora @PhilipNora7]

4.8 is coming. What sort of jump are we expecting and hoping for?

[Elon Musk @elonmusk]

We have dumped JAX/SGLang and rewritten the entire RL and inference stack from scratch in C/C++, which took some extra time.

Training (aka pre-training) is done. Lot of cleanup was needed on the SFT data to get that mostly right. Now, 4.8 is in RL and the evals are starting to climb. We will do a soft rollout to Bot users later this week and see how it goes.

Grok 4.8 will bear no resemblance to 4.7 (thankfully). It is not yet incredible, but, this time, the foundations are built right, so I expect it to evolve to be something special.

[Philip Nora @PhilipNora7]

@Da7_Tech @cognition How are you planning on utilizing carbon? Looks like it is best for security and architecture design, I’m curious how it will do in research compared to Argon

[Philip Nora @PhilipNora7]

@bot find me the best/highest value chinese X posts/accounts possible based on my interest in ai

[Elon Musk @elonmusk]

Chinese 𝕏 is 牛逼

[Philip Nora @PhilipNora7]

@Da7_Tech Qwen4 is probably our best hope. Even then idk

[Philip Nora @PhilipNora7]

@Da7_Tech This might be a stupid question but if grok bot is using opus 5.5 for certain tasks, will this mean that they might tap into fable at all? Even if its 5.1 it would be cool. But then again if we’re having these sorts of usage issues Fable definitely wouldn’t help its case

[Philip Nora @PhilipNora7]

@Da7_Tech They upgraded the usage so its good but the ram issue is still there in the desktop app

[Philip Nora @PhilipNora7]

@Da7_Tech Holy shit. Thank you

[Philip Nora @PhilipNora7]

@ConnorTalksAI Models get the general audience excited. I couldn’t agree more with you though, there’s times where a good harness can take a mid model and make it much much better