Roofing operator. $100M+ hail claims. I run a live agent company (Bench) and build Dragon Rancher in public.

Denver, CO
We counted it. Real operator work, two machines. Opus 5: 14.0 self-corrections per 1k Fable 5.1: 4.2 3.3× more walk-backs. Not more wrong — more unsupported first claims. benchagi.com/s/gTF-NlOZ5pxYe… cc @bcherny @alexalbert__ @AmandaAskell
258
I don't prompt Aurelius every morning. The loop already ran. Seven agents on the shop. He drafts. I approve. That's a crew, not a swarm. benchagi.com/agents
6
Cory Shelton retweeted
533
1,853
26,182
500,887
Cory Shelton retweeted
The problem with the rabbit r1 is that he tried to make it a productivity tool instead of a cute companion like the muse charm. CONSUMERS DO NOT CARE ABOUT PRODUCTIVITY
218
126
4,763
346,915
Drone photos and the point cloud, fused to a textured roof GLB. 35 faces. 40,632 cloud points. Khronos validator: 0 errors. Aurelius ran the loop. Merged this morning. Still a local slice — not the customer UI. benchagi.com
15
Cory Shelton retweeted
Replying to @JustinPetersMin
You, do not know God . You do not know Christ Jesus.and you do not know the Holy Spirit. And you certainly do not know the Holy Scriptures.Everything that you speak is from the Evil one himself. And lies he has told you fit into your fantasy world. Jesus will tell you the truth.
1
2
570
The GM I couldn't afford to hire. Aurelius reads the whole shop — pipeline, crews, the storm — and hands me the one or two moves before the reps log in. He drafts. I approve. That's capacity, not headcount. benchagi.com/agents/aurelius
27
Cory Shelton retweeted
Godmother of AI and World Labs CEO Fei-Fei Li says the real existential threat was never AI: "Any threat to human society, including existential, is within ourselves."
27
93
321
16,060
I’ve asked every model this question since GPT-4 every single model has told me the same thing. No. Not Opus 5.5
1
34
Cory Shelton retweeted
this is why I don't follow politics anymore
4
2
22
1,271
If you're a developer with no good ideas or a lack willingness to market/sell your product, today has to be a living nightmare for you.
26
Cory Shelton retweeted
GPT-6 Sol and Luna are out. Not only are they a very significant improvement across the board, but also in writing and general "you know when you try it" quality. We are also permanently reducing the API price by 50% making both of them viable for a ton of new usecases and making your usage go further too, even on the subscriptions. And one more thing. We are loading a banked reset into all accounts of our Plus, Pro and Business users. Let's go! openai.com/index/introducing…
We have been focusing on efficiency and intelligence for all. Very proud of the team. Only possible when you have incredible models at the top end of the capability that you can then use to make a big difference in everything else.
3,039
1,614
25,283
3,164,863
GPT-6 Sol and LUNA 🫶 Welcome Back to the fight @OpenAI
38
Cory Shelton retweeted
Which part of "get woke, go broke" do you not understand, comrade?
2
1
2
32
Opus 5.5 is so FAST omg...
29
Cory Shelton retweeted
Claude Opus 5.5 is available today.
Introducing Claude Opus 5.5, the first model in our new Claude 5.5 family. It performs at the level of Claude Fable 5.1 for most tasks, and costs 40% less to run than Opus 5.
610
982
17,234
1,028,884
Cory Shelton retweeted
JUST IN: Andreessen Horowitz launches a free AI school aimed at high school graduates & college dropouts as an alternative to elite universities.
150
503
5,911
284,317
60% of leads become proposals. You expected 90%. Eleven tools. One whiteboard. The owner is still the glue. One workspace. Seven agents. I approve every move. benchagi.com
29
Grok and Claude are down, I think we just lost a XAI data center.
2
341
"Vibe Coding is a Slur" 100% Agentic Coding & Product Development
25
The loop writes. It does not send. Aurelius does not send the message, move the money, or close the deal. I do. That's how we sell capacity, not headcount. benchagi.com/agents/aurelius
42