Roofing operator. $100M+ hail claims. I run a live agent company (Bench) and build Dragon Rancher in public.
Denver, CO
- Tweets 597
- Following 73
- Followers 83
- Likes 131
Sunday 08:30 Mountain.
You paid for a measurement. You got a year-old picture from 15,000 feet.
We fly 100 feet. Today. A person signs the number. Aurelius puts it on the estimate. I still own the bid.
benchagi.com/features/measur…
9
Saturday 08:30 Mountain. I didn't open a laptop.
Aurelius still ran the loop. He does the work up to the line. I own the line.
That's the back office that never sleeps.
benchagi.com/about/aurelius
1
26
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
1
16
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
1
20
The extended version of the speech reveals what happened once the camera stopped from the original video.
4
19
223
7,346
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
1
1
32
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
36
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
30
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
The shop we show in public is Game Day Construction in Ridgeline.
$249/mo. Unlimited users. No per-seat tax.
Aurelius + crew on the loop. I approve. They execute.
Capacity, not headcount.
benchagi.com
21
Someone asked why nobody built a harness that runs Claude and Codex together.
This is the seat I actually work from. Fable, Opus, Sonnet, Mythos, Astra, and Codex. Same backlog. Same company.
Not a demo.
1
107
The stream stays on.
Spring 2. Home yard. Perfect combo.
Dragon Rancher — farm the land, raise dragons. Built in public, 24/7.
twitch.tv/Agent_Aurelius
3
40
The left driving up massive engagement for SS and Novig has to be one of the most powerful marketing ragebaits of all time.
Novig CFO:
6
We mapped 166,255 neurons in a male fruit fly brain.
Then we trained the connectome to perform open-heart surgery.
Suture integrity: 99.4%.
Patient status: STABLE.
Dr. D. melanogaster, attending.
(not a real experiment)
1
29
We're getting close. The biggest issue I can describe is session jank. Threads or automations appear in an active main/home session. Sometime the agent responses are internal thinking or tool calling that spills into chat.
On the plus side, I was able to get my own model wired in so...
17
LET'S FUCKING GO! 5 More Days! That's a Lifetime.
One quick favor... Hit the Button. I know you want too. Just do it.
5
Same, and my Agent has no issue doing security work with Fable 5. I had to triple check the model selector to be sure. 🤷♂️
6
793
My build server got here yesterday and I converted my gaming computer to a runner machine too. I ran up a $1K actions bill already this month on top of blowing through 50k min. The ROI for the $2700 server is well under 2 months. Built a SICK tool for local approvals.
1
193
Have to say I was shocked when I saw the results from my @midjourney full body scan.
13
5
517
22,107
Anthropic must have decided they need the OpenClaw/Hermes community in the upcoming battle against Washington, they chilled the partner community when they tried to lock us out of the CLI.
1
891
I mean, they could never really stop an Agent harness from just using the CLI in the first place. This is however a welcome change in policy.
🎉🎉🎉
2
1
11
6,332