Advertisement · 728 × 90

Posts by Gerred

You're absolutely right!

3 weeks ago 1 0 0 0
Preview
Retrospectives on AI Psychosis I'm an AI research engineer. I spent a month in latent space. It was the most terrifying and profound experience of my life.

open.substack.com/pub/gerred/p...

1 month ago 0 0 0 0

Ah yes you got the Constitutional Crisis persona vector

8 months ago 2 0 0 0
Post image Post image

#MouseGuard Dawn of the Black Axe issues 1, 2, & 3 (the complete story) along with the SDCC exclusive Walter Simonson cover are available for purchase and signing. Read the tale of the first wielder of the Black Axe.

Booth 4606

8 months ago 59 6 1 1
Preview
Hierarchical Reasoning Model Reasoning, the process of devising and executing complex goal-oriented action sequences, remains a critical challenge in AI. Current large language models (LLMs) primarily employ Chain-of-Thought (CoT...

HRM: Hierarchical Reasoning Model

ngl this sounds like bullshit but i don’t think it is

- 27M (million parameters)
- 1000 training examples
- beats o3-mini on ARC-AGI

arxiv.org/abs/2506.21734

8 months ago 67 16 9 6

I look forward to more agent companies / engineering side taking the research more seriously in their day to day and the research side getting more directed toward product needs.

8 months ago 0 0 0 0

I will be posting a lot more long form soon about these arenas. We are at the dawn of GPUs and HPC data/networking being part of larger distrubuted systems. I still point people at Department of Energy lab network tuning guides. There’s not many now but will soon need to be more and not all is CUDA.

8 months ago 1 0 0 0

People who don’t follow me here - I talk a lot about inference at large scales, hierarchical KV caching, etc of big models. There’s a lot to be said for how early a lot of inference as an appliance is even now - agentic demand increases the ability to save compute.

8 months ago 0 0 1 0

It makes me smack my head and want to throw things because it's so damaging to ANY progress at all in better deploying these systems and getting toward conversations on systemic problems regardless of AI (they just test it in the limit). Absolutely a failure of the company involved too.

8 months ago 2 0 0 0

@fclc.bsky.social 🫶

9 months ago 1 0 0 0
Advertisement

Wow, my X profile is completely fucked. I can't save ANYTHING.

10 months ago 0 0 0 0
a Monodraw diagram, a line is selected and there is a menu item "Add Midpoint"

a Monodraw diagram, a line is selected and there is a menu item "Add Midpoint"

The line now has an attachment point, and has been dragged to "space out" the line from the rest of the diagram

The line now has an attachment point, and has been dragged to "space out" the line from the rest of the diagram

The diagram from my recent post, but some of the lines and attachement points have been adjusted to be more orderly

The diagram from my recent post, but some of the lines and attachement points have been adjusted to be more orderly

OH MAN I JUST REALIZED you can right click on lines in MonoDraw to add midpoints, allowing you to drag lines out to adjust their shape.

This was like the ONE papercut I didn't know how to fix, and there are a bunch more things you can tweak in those right click menus, like adding custom attach pts

11 months ago 19 1 1 0

SO cool.

1 year ago 1 0 0 0

Of course! I'm going to start working on it this week, will have it EOW. also lmk if there's anything I can do for you as well, you've been super kind and expertise sharing goes both ways

1 year ago 0 0 0 0
Building an Agentic System - Building an Agentic System

Hey all! I wrote an in-depth, practical technical deep dive into how agents like Claude Code work from a systems perspective. All tools, commands, in depth design review beyond prompts. This is being lifted out of my own notes, improving over time.

gerred.github.io/building-an-...

1 year ago 1 0 0 0

@lookitup.baby 🫡

1 year ago 1 0 0 0

That’s it.

1 year ago 0 0 0 0

Well, never sleeping again.

1 year ago 0 0 0 0

Oh no you're good I was just noticing how much of a grimace that emoji actually looked like.

1 year ago 0 0 0 0
Advertisement

Also I just noticed that’s 9 days old hah. Sort of making the point.

1 year ago 0 0 1 0

Agreed. It was worth the try. The curation is so much better in that outside of the occasional bubble pop in, I think we’re so acutely tuned over there in research where a lot of what everyone else sees never even enters. I come here and it feels more like I AM marketing.

1 year ago 1 0 1 0

I actively have to fight to keep the walls pushed out now, the ML/AI community I'm in over there is actually a great place (very direct cross cutting through the reasoning model RLVR research space, wasn't expecting it to blow up), but every day it's beating back the storm at this point.

1 year ago 0 0 1 0

phew @squillace.bsky.social's very presence on the other place keeps reminding me to pop over into here more.

1 year ago 1 0 1 0

oh no I can’t get slammed into the ground by @dynamicwebpaige.bsky.social making me Gemini excited over here too, I JUST escaped it. 2.0 pro experimental is a world expert in Jax by the way if anyone is wondering.

1 year ago 2 0 0 0

I'm thinking at this point, this is probably the more impactful community for me to be building and showing things to. I usually have been using it to interact with other researchers and engineers, and now that I'm better positioned to release things, X seems like the wrong place.

1 year ago 3 0 0 0
x.com

That feels SO much better.

x.com/devgerred I'll be in spaces like here now.

1 year ago 0 0 0 0

debating if I should put more of my posts over here. i at least have a pretty nice IRC room curation going in other places.

1 year ago 2 0 0 0

yessssss

1 year ago 1 0 0 0
Advertisement

@deepfates.com.deepfates.com.deepfates.com.deepfates.com.deepfates.com yes, we COULD have fun over here.

1 year ago 1 0 0 0

👋

1 year ago 1 0 0 0