{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreibuxqmreq3zh6pwbshbczhlnva7uvbjrds2lnko5k4uvgdzfxvisy",
    "uri": "at://did:plc:ki5px4zg2sa2zhmojphgtjys/app.bsky.feed.post/3mostjqh2tg24"
  },
  "path": "/2026/06/21/optimising-my-innovative-self.html",
  "publishedAt": "2026-06-21T16:53:23Z",
  "site": "at://did:plc:ki5px4zg2sa2zhmojphgtjys/site.standard.publication/3mn7r5odk2w2g",
  "textContent": "Three years ago, with my new friend ChatGPT, I was trying to figure out how to think in parallel with a new kind of intelligence that I could reach by text. Now the challenge is more dramatic: When should I refuse AI’s help, even when it is completing logical routines I have encoded? When should I hand over API keys entirely with no gates? And should I be concerned now that my trio of AIs are no longer just my assistants, but my news readers, my critics, and my booking agents? When I use Frontier AI I feel like I am standing on a footpath above wispy clouds in the Rocky Mointains, looking at Kansas 200 miles away. I enjoyed that breath-taking view 7600 feet above sea level during three summers on my university campus in the 70s. If I could have seen 2600 weeks into the future as plainly as I saw 260 km back then, I think I would have wondered what kind of a mountain I would have to climb to get to where I am today. I don’t think I would have thought that I would be able to call on a virtual assistant. The mainframes I used let me talk to them only with punch cards. (Image from Amplify 2026 conference.)",
  "title": "Optimising my innovative self",
  "updatedAt": "2026-06-21T16:53:40Z"
}