Back in the '90s, running a small record label, I had to learn Excel, Filemaker Pro, Pagemaker, even some Photoshop, to deal with mechanical rights, music publishing, artist contracts, distributor agreements, customer service, marketing, album design, printing, manufacturing, etc. Now it sounds as though I could just learn Agentic IA. But would I still understand the nuts and bolts of what was happening?
You can instruct the agents to give you the level of detal about their actions that makes you comfortable. The key skills to have are specification and validation. You have to know enough about what you want done to be able to give the agents unambiguous instructions, and you must validate the results.
There's a story going around that a programmer instructed Claude to take his company's data and produce reports. It was a few months later that someone realized that Claude was fabricating its own data for the reports. Don't let that be you.
Autonomy is racing ahead of security. Be very careful, and if you don't know what you're doing, don't. Agents are going to be made safer, wait a few months to jump in.
My concern about Agents is the potential for hacking and how that agent could be turned against you in some pretty terrible ways. Do you see that as a legitimate risk concern?
Every generation gets one asymmetric window—railroads, oil, the internet, mobile. Most people miss it because it looks messy before it looks obvious. Agentic AI may be that window. If tools like OpenClaw truly collapse execution costs—research, outreach, automation, workflow—then leverage shifts from capital to competence. But here’s the hard truth: technology doesn’t reward spectators. It rewards builders. The same systems that empower entrepreneurs will also displace complacent professionals. Mastery is optional. Consequences are not. If execution becomes cheap, judgment becomes king. Learn the systems. Control the agents. Or compete against someone who did.
Extending this thought process. Agents will conduct commerce - how much of that in Stablecoin for instant transfer and settlement? Implications for USDC, USDT, BTC?
My concern is for replacing the human work force. Streamlining is essential but it always eliminates jobs.
Thank you for this!
Back in the '90s, running a small record label, I had to learn Excel, Filemaker Pro, Pagemaker, even some Photoshop, to deal with mechanical rights, music publishing, artist contracts, distributor agreements, customer service, marketing, album design, printing, manufacturing, etc. Now it sounds as though I could just learn Agentic IA. But would I still understand the nuts and bolts of what was happening?
You can instruct the agents to give you the level of detal about their actions that makes you comfortable. The key skills to have are specification and validation. You have to know enough about what you want done to be able to give the agents unambiguous instructions, and you must validate the results.
There's a story going around that a programmer instructed Claude to take his company's data and produce reports. It was a few months later that someone realized that Claude was fabricating its own data for the reports. Don't let that be you.
One word, DOS.
Autonomy is racing ahead of security. Be very careful, and if you don't know what you're doing, don't. Agents are going to be made safer, wait a few months to jump in.
Brilliant as always.
My concern about Agents is the potential for hacking and how that agent could be turned against you in some pretty terrible ways. Do you see that as a legitimate risk concern?
Security is awful right now, but will get better.
Every generation gets one asymmetric window—railroads, oil, the internet, mobile. Most people miss it because it looks messy before it looks obvious. Agentic AI may be that window. If tools like OpenClaw truly collapse execution costs—research, outreach, automation, workflow—then leverage shifts from capital to competence. But here’s the hard truth: technology doesn’t reward spectators. It rewards builders. The same systems that empower entrepreneurs will also displace complacent professionals. Mastery is optional. Consequences are not. If execution becomes cheap, judgment becomes king. Learn the systems. Control the agents. Or compete against someone who did.
Extending this thought process. Agents will conduct commerce - how much of that in Stablecoin for instant transfer and settlement? Implications for USDC, USDT, BTC?
It's almost all going to be in crypto, and mostly in stablecoins.