Greg Kamradt
@GregKamradt
Followers
40K
Following
14K
Media
934
Statuses
7K
President @arcprize, Founder https://t.co/clDqzdHe1A, builder/engineer
San Francisco, CA
Joined January 2011
RT @mikeknoop: Zooming out ARC progress: I'd say OpenAI's o-series progression on v1 is a bigger deal than Grok's progression on v2 (so far….
0
11
0
We got a call from @xai 24 hours ago. “We want to test Grok 4 on ARC-AGI”. We heard the rumors. We knew it would be good. We didn’t know it would become the #1 public model on ARC-AGI. Here’s the testing story and what the results mean:. Yesterday, we chatted with Jimmy from the.
Grok 4 (Thinking) achieves new SOTA on ARC-AGI-2 with 15.9%. This nearly doubles the previous commercial SOTA and tops the current Kaggle competition SOTA
231
661
6K
Next week we're hosting a developer preview of ARC-AGI-3. Open to sponsors, donors, and reseachers. I'll announce 4 things to launch (ARC Prizes first infra 👀). Then we'll have @dwarkesh_sp do a live fireside with @fchollet . Come join (few public slots left).
ARC-AGI-3 Developer Preview . * Hands on first look at ARC-AGI-3 (live demos & API access).* Fireside with @fchollet moderated by @dwarkesh_sp. 7/17, San Francisco. Open to sponsors & researchers of @arcprize (very limited public slots available)
5
5
59
Of course this isn't an official study, but directionally o3 does great. Chat thread.
1
0
7
Grid > Object Detection. Did a great job qualitatively describing the input task. I wasn't very generous with it either (I didn't supply the output)
1
0
7
Image > Grid. I asked o3 to label each color, it did so just fine. You can use json grids or images for your submission. So there isn't a problem identifying colors if you're using vllms
1
0
6
I hear "models don't do well on @arcprize because they can't perceive" all the time. What test would disprove this?. I tried 3 tests:.1. Image > Grid.2. Grid > Image.3. Grid > Object Detection. o3 got them all right - results:
4
3
54
@adcock_brett what would be a good activity to make robots compete at? A good nonviolent spectator sport.
1
0
2
RT @typewriters: next week! . I ascribe to the @arcprize definition of AGI - as long as there are tasks that are easy for humans and hard (….
0
3
0
the todo/md and plan/md paradigm will cross-transfer to non-coding domains. a clear plan is clear thought. it's a horizontal skill that covers nearly 100% of work.
btw by far the most interesting new pattern we are seeing in @latentspacepod (upcoming with @olivierddr and @sdrzn and @pashmerepat) . is that people are using @claude_code and @ClineBot for non-coding tasks and this is becoming surprisingly effective for things like sales/BI
1
0
3
.@PabloMotoa has helped me with editing for a while, now he is digging into his own content. This one is on @levelsio - very cool!.
most people know @levelsio as the indie hacker and entrepreneur making over $200k/month. but what most people don’t know is that he:. - was a successful youtuber.- nearly went broke multiple times.- built his personal brand by posting 10 times a day. he’s one of the
2
0
15