AgenticRebirth

Building with AI || Pi maximalist and benchmaxxer || Experiencing the agony of insufficient GPUs

[AgenticRebirth @AgenticRebirth]

@goodalexander Which provider did you end up landing on?

[AgenticRebirth @AgenticRebirth]

@usr_bin_roygbiv I’m beyond jealous… I assume at some point we’re going to have a giant memory glut but it’ll be years away.

[AgenticRebirth @AgenticRebirth]

@TheAhmadOsman - one RTX 3090 or one RTX Pro 6000? Does 96gb vram actually unlock anything drastically better?

[AgenticRebirth @AgenticRebirth]

Unfortunately this account has been shadowbanned for 10 days, for no reason. All attempts to request support from X have been ignored. If you can see this and are willing to ping @nikitabier, I would appreciate it. If this account is not restored soon, I will have to leave.

[AgenticRebirth @AgenticRebirth]

@mvanhorn Try using orca. It will preserve your sessions.

[AgenticRebirth @AgenticRebirth]

I have some interesting benchmarks on AI agent harnesses to post, but unsure if anyone can see my posts. If you can, could you please interact below?

[AgenticRebirth @AgenticRebirth]

@karpathy @gallabytes It's not obvious how this materially differs from a 'claw'. Can you elaborate on what you think makes this such a giant leap compared to tagging your agent in a Discord or Telegram chat?

[AgenticRebirth @AgenticRebirth]

While I like omp a lot in concept, I (sadly) still find codex to be more reliable. Even so, I will keep using/trying to propose improvements to omp since I don't want my model and my harness to be bound together.

[Rhys @RhysSullivan]

people haven't used the codex desktop app and it shows

[AgenticRebirth @AgenticRebirth]

this has been great marketing for fable, whole tl can't wait to get back to using it. they'll have 5x the demand on relaunch as they had on launch.

[AgenticRebirth @AgenticRebirth]

@omarsar0 There has actually been some work on this, I’m not surprised to see this approach works

https://t.co/RLlklTadXB

[AgenticRebirth @AgenticRebirth]

We still know remarkably little about how to extract maximum performance from current models. The research on this topic seems extremely inconsistent, or at least difficult to interpret cohesively. Nevertheless, this is very cool.

[OpenRouter @OpenRouter]

Introducing the Fusion API, the smartest compound model in the market.

Fusion achieves Fable-level intelligence at half the price.

How it works 👇

[AgenticRebirth @AgenticRebirth]

@sudoingX they're going to work closely with USG to make sure it's cleared after the anthropic debacle - which means it's probably going to be nerfed compared to what it could be, but still an improvement

[AgenticRebirth @AgenticRebirth]

if there's an economic downturn and people start dumping their used GPUs you bet your ass I'm slurping those up

[AgenticRebirth @AgenticRebirth]

This isn't new/surprising - in theory I knew it - but somehow it's still surprising to observe.

[AgenticRebirth @AgenticRebirth]

An observation from doing multiple terminal-bench runs is that results can vary significantly even if you change nothing. LLMs just don't work deterministically: if you run a test 5 times, they can fail twice and pass 3 times despite being configured exactly the same way.

[AgenticRebirth @AgenticRebirth]

@DavidOndrej1 I aint sleeping

https://t.co/qAKbKLiiex

[AgenticRebirth @AgenticRebirth]

Baseline: Pi with GPT-5.5 (medium) scores 70.8% on Terminal Bench 2.1 at a cost of $35.14.

Next: see if we can improve performance by tuning only the system prompt, without increasing cost.

[AgenticRebirth @AgenticRebirth]

Baseline: Pi with GPT-5.5 (medium) scores 70.8% on Terminal Bench 2.1 at a cost of $35.14.

Next: see if we can improve performance by tuning only the system prompt, without increasing cost.

[AgenticRebirth @AgenticRebirth]

Today's Day 1: Deep-diving into agent harnesses. Let's find out what it takes to squeeze max performance from frontier models.

Goal: Best quality + speed at the lowest token cost.

Inspired by @usr_bin_roygbiv’s cheerleading, I’m testing with the Pi harness and optimising 3 core dimensions:
1. Context management.
2. Tools.
3. Control logic (loops, workflows, determinism).

The loop is simple:
• Run baseline benchmarks in Pi
• Generate a narrow hypothesis on one dimension
• Test and measure
• Iterate

Hoping to ship a stronger harness + learn a ton along the way. Follow along for updates and results!

What harnesses are you running that I should try learn from? Drop your suggestions below 👇

[AgenticRebirth @AgenticRebirth]

@usr_bin_roygbiv Don’t fucking tell me that I already hate myself for not committing sooner. I don’t even wanna look at the ram I’m gonna have to buy

[AgenticRebirth @AgenticRebirth]

Extremely plausible. I’m buying an RTX Pro 6000 today.

[Ahmad @TheAhmadOsman]

Next up is ID verification for AI models btw

"...suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees."