@PromptLLM This is Robert Fritz's "structural tension" from 'The Path of Least Resistance' repackaged.
Although Fritz points out you can't "strip" the emotional weight by deciding to. The result has to stay charged or nothing pulls.
@NicerInPerson I dunno. I don't think the general populace is going to let agents go YOLO with their credit cards for quite some time.
It will be fun finding out over the next few years though :o)
I think services that make money from upsell and inertia will start using Apple's App Attest / Google Play Integrity to restrict direct API access by personal AI agents.
My agent wants to find a better deal. They want to keep me in their app.
@NicerInPerson There might be middle road, where agents are permitted to load baskets / get to the booking screen, and companies can still do upsell / on sell / comparison, and confirm with the end user it’s what they want.
Agent books restaurant and OpenTable push notify me to confirm?
I do like the tiling hyprland window manager in Omarchy, so I added some features to Aerospace (better multimonitor support, focus mode, better single window handling) and that is now my daily driver on my macs.. Love it. https://t.co/pvOZFhIOzy
@stevegraham With those constraints that's crazy!!
I went trial and error, but Claude did build me CLIs for Things (fave task app) and Strong (fave weight training app). Both have cloud sync, but no public API, and now agents run em both.
@stevegraham I’m continually reminded of Neal Stephenson’s “The diamond age”. There should be a charity to give motivated clever inner city kids Pro/max accounts.
@NicerInPerson@jonas@Pebble I made an apple watch complication / app for that collection (day note / task / idea). a local agent at home deals with the details.
You wonder how many people have ever lived, so you ask an agent to build you an interactive exploration while you do other things.... crazy times
https://t.co/wlJLuA6U1Z
I asked @claudeai Opus 5.5 to make an animated video about a robot who realises he's stuck in an AI-generated world.
I wanted to really push it: 12 worlds, 12 styles, from 8-bit to claymation to pencil sketch, you can really see how well this model has imagined all worlds.
This is the best animated video you will see today made by Opus 5.5, I promise.
i'm still not sure people are fully getting how good opus 5.5
i gave my claude agent fig access to the runway MCP and told it to make a 'high-end netflix style documentary about superintelligence for normies' and...
it came back with this
Tested lots of research tools today. (you .com, parallel, perplexity, etc.)
most interesting finding: Stating exclusions as a deliverable (“A section with sources considered and rejected”) took compliance from leaky to 100% where the same rule as an instruction was ignored. 🤷🏻
@NateKettles@SpencerGBull Thanks again. I just had a Claude /deep-research query spend 5.66M tokens in 27 minutes across 106 agents, so I'm feeling that pain already. 😬
I got inspired by https://t.co/mXKupIDCoA @SpencerGBull - and repurposed an old iPad.
It tracks agents in Herdr, and lets me switch between them with a tap of the screen. The game changer was getting a local agent to title and subtitle the agents using their transcripts.
Our new Dell 52" 6k Ultra Wide feels like a full wall of display. Works great in Omarchy over a single thunderbolt cable for charging and display. But I still can't ditch the multi-agent command center bottom screen to help manage herdr while I multi-task.
@NateKettles@SpencerGBull Thanks for the tip.
I'm researching for a book I'm writing and I've been testing perplexity deep research and model council vs claude deep research vs chatGPT with a number of prompts and effort levels.