{
  "title": "Best Practices in AI Work | Anssi Nurminen, Lasse Mikkonen | Neuvottelija 408",
  "titleOriginal": "Tekoälytyön huippukäytännöt | Anssi Nurminen Lasse Mikkonen | Neuvottelija 408",
  "episodeNumber": "408",
  "guest": "Anssi Nurminen, Lasse Mikkonen",
  "publishedAt": "2026-09-17",
  "duration": "1:01:45",
  "durationIso": "PT1H1M45S",
  "youtube": "https://www.youtube.com/watch?v=UaG4uuMYDi8",
  "fiCanonical": "https://www.neuvottelija.fi/fi/episodes/854-tekoalytyon-huippukaytannot-anssi-nurminen-lasse-mikkonen-ne/",
  "originalLanguage": "fi",
  "format": "full",
  "topics": [
    "ai_enterprise_tech"
  ],
  "description": "Anssi Nurminen and Lasse Mikkonen compare their own AI workers with the host's and ask what agentic work actually requires. Nurminen's agent Anneli turns WhatsApp photos, video or plain text into conceptualised videos and updates its skills from feedback; Mikkonen's Ossi and Dude reconcile receipts, keep a case memory and read email. The episode defines MCP, skills and the harness, sets out two memory architectures, and keeps three disagreements open: own hardware or cloud, whether security matters for hobby use, and whether the learning jump can be made in one go later.",
  "chapters": [
    {
      "t": "00:00",
      "title": "Guests Anssi Nurminen and Lasse Mikkonen",
      "seconds": 0
    },
    {
      "t": "00:09",
      "title": "The AI working group and its 14 AI workers",
      "seconds": 9
    },
    {
      "t": "00:48",
      "title": "Anssi's background and the Anneli agent",
      "seconds": 48
    },
    {
      "t": "01:31",
      "title": "Lasse's path from DevOps to agentic work",
      "seconds": 91
    },
    {
      "t": "02:33",
      "title": "Ossi and Dude handle the receipt drudgery",
      "seconds": 153
    },
    {
      "t": "03:51",
      "title": "How Anneli makes videos from WhatsApp",
      "seconds": 231
    },
    {
      "t": "05:12",
      "title": "Teaching Anneli through feedback",
      "seconds": 312
    },
    {
      "t": "06:32",
      "title": "What MCP actually is",
      "seconds": 392
    },
    {
      "t": "06:56",
      "title": "Skills for an AI worker",
      "seconds": 416
    },
    {
      "t": "07:46",
      "title": "Video as a way into the restaurant trade",
      "seconds": 466
    },
    {
      "t": "08:48",
      "title": "Meta Ray-Ban glasses in use",
      "seconds": 528
    },
    {
      "t": "10:05",
      "title": "Image search through the glasses camera",
      "seconds": 605
    },
    {
      "t": "10:39",
      "title": "Digital footprint and anonymity",
      "seconds": 639
    },
    {
      "t": "11:57",
      "title": "Making videos while wearing the glasses",
      "seconds": 717
    },
    {
      "t": "12:38",
      "title": "Continuous recording and strengthening memory",
      "seconds": 758
    },
    {
      "t": "13:47",
      "title": "Finnish is not supported in the glasses",
      "seconds": 827
    },
    {
      "t": "16:21",
      "title": "Why augmented reality is the more interesting one",
      "seconds": 981
    },
    {
      "t": "17:07",
      "title": "Video models and China's lead",
      "seconds": 1027
    },
    {
      "t": "18:21",
      "title": "Finland's platform-economy debate is behind",
      "seconds": 1101
    },
    {
      "t": "19:07",
      "title": "Is it worth waiting for a laid table",
      "seconds": 1147
    },
    {
      "t": "21:34",
      "title": "Vibe coding versus real competence",
      "seconds": 1294
    },
    {
      "t": "23:19",
      "title": "Delegating to agents and what breakage teaches",
      "seconds": 1399
    },
    {
      "t": "24:47",
      "title": "The harness keeps the agent on track",
      "seconds": 1487
    },
    {
      "t": "26:19",
      "title": "When an agent leaked its reasoning into the group",
      "seconds": 1579
    },
    {
      "t": "27:49",
      "title": "What other skills have been built",
      "seconds": 1669
    },
    {
      "t": "28:39",
      "title": "Dude's case memory and meeting notes",
      "seconds": 1719
    },
    {
      "t": "30:23",
      "title": "A restaurant profitability skill as a joint project",
      "seconds": 1823
    },
    {
      "t": "31:01",
      "title": "Long-form video and character consistency",
      "seconds": 1861
    },
    {
      "t": "33:28",
      "title": "Silicon Valley's sense of time versus Europe",
      "seconds": 2008
    },
    {
      "t": "35:04",
      "title": "Why it is worth trying at any level",
      "seconds": 2104
    },
    {
      "t": "36:19",
      "title": "Security and the Red Queen effect",
      "seconds": 2179
    },
    {
      "t": "39:03",
      "title": "Mac Mini, Nvidia and local compute",
      "seconds": 2343
    },
    {
      "t": "41:13",
      "title": "Pseudonymisation before sending to the cloud",
      "seconds": 2473
    },
    {
      "t": "43:00",
      "title": "Europe's second chance in models",
      "seconds": 2580
    },
    {
      "t": "44:21",
      "title": "A monthly subscription is a cheap investment in yourself",
      "seconds": 2661
    },
    {
      "t": "44:56",
      "title": "Taking an agent to the company IT department",
      "seconds": 2696
    },
    {
      "t": "47:05",
      "title": "The Hello Humans podcast simulator",
      "seconds": 2825
    },
    {
      "t": "48:29",
      "title": "Crowdsourcing and trading skills",
      "seconds": 2909
    },
    {
      "t": "50:03",
      "title": "Memory structure as the year's main theme",
      "seconds": 3003
    },
    {
      "t": "52:00",
      "title": "Graphify and a custom case template",
      "seconds": 3120
    },
    {
      "t": "54:35",
      "title": "Memory already works in ChatGPT and Claude",
      "seconds": 3275
    },
    {
      "t": "56:13",
      "title": "Skill updates based on the conversation",
      "seconds": 3373
    },
    {
      "t": "57:01",
      "title": "A Finnish-language skill removes the slop",
      "seconds": 3421
    },
    {
      "t": "57:50",
      "title": "Bots talking to each other in a liminal state",
      "seconds": 3470
    },
    {
      "t": "58:48",
      "title": "Kesko's AI strategy and training",
      "seconds": 3528
    },
    {
      "t": "59:50",
      "title": "The race between America and China in models",
      "seconds": 3590
    },
    {
      "t": "1:01:15",
      "title": "Closing words and Sisäpiiri",
      "seconds": 3675
    }
  ],
  "sourceTranscriptType": "none",
  "subtitleMethod": "none",
  "provenance": "Owner page assembled from YouTube metadata, the neuvottelija.fi episode record and the publisher's own chapter marks, translated one for one. The summary was written from the uploader's own Finnish caption track. No English caption track exists, so no transcript is published here. Disclosure carried from the episode: the host sits on the board of Fredman Group, where Kespro, Anssi Nurminen's employer, is a partner.",
  "slug": "ep408-tekoalytyon-huippukaytannot-nurminen-mikkonen",
  "guestsCanonical": [
    "Anssi Nurminen",
    "Lasse Mikkonen"
  ],
  "guestPages": [
    "https://www.neuvottelija.com/podcast/guests/anssi-nurminen/",
    "https://www.neuvottelija.com/podcast/guests/lasse-mikkonen/"
  ],
  "page": "https://www.neuvottelija.com/podcast/episodes/ep408-tekoalytyon-huippukaytannot-nurminen-mikkonen/",
  "markdown": "https://www.neuvottelija.com/podcast/episodes/ep408-tekoalytyon-huippukaytannot-nurminen-mikkonen/index.md",
  "captions": null,
  "transcript": null,
  "videoId": "UaG4uuMYDi8",
  "dbId": null,
  "qaCoveragePct": 0.0,
  "qaSemanticVerdict": "metadata_only"
}