先说结论

TWIML AI Podcast 主持人 Sam Charrington 这次请回老朋友 Dan Jeffries（依据英文转写整理，全部 27 段、0s–3211s 已完整读取；两人上次对话是 2024 年底谈 computer use agents）。过去六个月 Jeffries 在造 Nautilo——他自称「30 个产品合一」的开源组织级 harness，一句话定位是「翻转盖手机版的 Her，但面向团队与家庭」（本站查证：Nautilo 已于 9 月开源发布，仓库在 agentsea/nautilo，官方帖对「秘密守护者」「命名空间」「时间限定解密委托」等表述与访谈完全一致）。整场对话最有价值的是三层判断：其一，个人 agent 刚跨过易用性门槛（节目举例 Meta 的 Muse、OpenAI 的 Dot 开箱即用），但按创新扩散曲线我们仍在 2% 的「早期采用者」段——享受 agent 红利的主要是本就有五十件事待办的高能动性人群，多数人面对无限画布不知道该干什么；其二，大量精心设计的多 Agent 组织架构图在他看来「基本是角色扮演（cosplay/LARPing）」——让同一个底层模型扮演的邮件 bot 和 CEO bot 互相开会，不是在做工作；其三，真正的机会在「有主见的设计」：隐私、自主权、协作共创这些通用大厂产品不会做的取舍。节目名「造 agentic 产品的残酷现实」在结尾兑现：Jeffries 与圈内创业者把当下创业环境称为「那堵墙（The Wall）」——建东西从未如此容易、被看见从未如此困难。

「经理人时代」：2% 的早期采用者与 10%/90% 饼图

Jeffries 用计算史给当下定位：这是八十年代末、九十年代初的个人电脑——「你默认它会出毛病、得知道怎么把它装回去」，而不是 iPhone 时代「你只管用它、坏了不知道该找谁」。他写过一篇《Agency Problem》：多数人不知道拿 agent 干什么，受益者集中在高能动性人群——「手头永远压着五件想创的业、三十本想写的书、五十个想去的 trip」的人，无限画布对他们是解放；对其他人则是一头雾水外加一点失业恐惧。他转推过一张 2026 年饼图自嘲：10% 的扇区是自己真正干活，90% 是在伺候那些替你干活的 agent——编程、review、迭代。Sam 用自己的体验背书：自动化确实爽，但那些东西经常坏，corner case 一箩筐，你总得回到代码里。Jeffries 补了个扎心的对照：一个懂技术的老朋友问他「我怎么入门 agent」，他发现这个问题对自己这类人根本不成立——「你跟它说话，它搞砸了你就再跟它说，想想哪错了」；但很多人要的是模板、是「告诉我怎么做的那个东西」，而行业还没有到给模板的阶段[120–360s]。

他引了一篇文章《Tool-Shaped Objects》：日本匠人的木工刨（kanna）是艺术品、刨木头如天使般精准，但经济上完全不值钱——电动刨以零头成本和零头时间干同样的活。他以此讽刺大量 AI 演示视频：「90% 的内容在搭环境，搭完之后真正用它干活的那段，不存在」。他管那些「我搭了邮件 bot、税务 bot，再搭个 CEO bot 管它们」的架构叫 faceless toaster bot（无脸烤面包机）与「LARPing 当 CEO」——「你的 agent 们在互相说话？它们说什么？为什么需要说？底下不是同一个模型吗？」[360–480s]

Nautilo 是什么：秘密守护者、命名空间与自主权

产品本体：Jeffries 把 agent 称为 Genie（精灵），定位是「老式意义上的 secretary——拉丁语 secretum，秘密守护者」：给每个用户一个深度人格化的超 agent，起名、配图、从三万个 ElevenLabs 音色里挑声音；它记住你的一切、从不 break character。团队重写了群聊：一个「魔法路由器」让人和 agent 在同一对话里自然说话、agent 被点名即醒，人机同为一等公民——之所以不直接接入 WhatsApp/Slack，是因为这些平台的条款诞生于「bot=垃圾信息」时代，接进去要么违规、要么只能挂一个没有人格连续性的单例 AI[600–720s]。

真正有主见的取舍在隐私。Jeffries 引 François Chollet 的判断「AI 将成为我们与世界的接口」，推到结论：十年后你可能都不记得怎么在手机上「打开一个应用」——而如果这个接口是一个 SaaS 应用，那就是「监控经济的平方」：你在它那里聊孩子在学校的问题、聊官司、聊体检结果，「社交媒体对社会造成的伤害，和这个相比不算什么」。他不指望立法解决，路线是赛博朋克式的自主权（self-sovereign）：开源全栈、自己控制服务端与客户端；端到端加密的变体——后端是分布式密钥，Genie 需要读明文，所以给它的是时间限定的解密委托：醒来、解密一条聊天或一段记忆、用完即停，没有常驻主密钥。另一层是命名空间（namespaces）：你跟自己的 Genie 一对一说的事，不会漏进有同事在的群聊——「节目举例 Muse 把某人的家庭住址说了出去，就是因为它什么都知道而没有命名空间」。他还给模型做隐私分级路由：敏感话题优先走 Venice 这类隐私通道[720–840s, 960–1080s, 2160–2280s]。

多 Agent 之争、记忆系统与「哑锤哲学」

Sam 直接唱反调：他用 Hermes（Nous Research 的个人 agent）单 agent 拉技能处理所有领域，「干得挺好」，凭什么要多 Agent？Jeffries 的回应值得整段记下：他也爱 Hermes 团队（转写提到 Teknium 等成员），而且哲学其实相同——他信的是一个无所不能的超强 Genie。「这是宠物 vs 牛群的问题：agent 可以自己去 spin up 牛群，但我不想认识牛的名字。」他反对的只是「烤面包机军团」；他说的多 Agent 是「人和 agent 在同一场对话里平起平坐」，这需要在后端造一堆小分类器和路由、共享哪些记忆等机制。一个具体场景：你让 agent 研究明年夏天的旅行，然后你变成「肉身信鸽」把结论复述给配偶——不如把带着记忆与报告的 agent 直接派去和你配偶聊，带回她的意见。Agent 的价值是「上下文守护者+秘密守护者」，人不该在没有完整上下文时当搬运工[840–1200s]。

记忆怎么做？「上下文窗口永远不够，它们全都是 Mr. Meeseeks（瑞克和莫蒂里任务完成即消失的临时角色）。」Nautilo 的方案是压缩哈希：每段记忆存元数据加一个「意义的线索」哈希，agent 先扫哈希、判断哪些值得展开，展开一小段再筛、选中才全文展开读，「读完再开口」；记忆随时间衰减；他称这套东西是记忆突破、打算写论文。一个他们自己也解释不了的观察：即便跑同一个模型，不同 Genie 在各自积累上下文后会呈现出「惊人的人格多样性」。而这套检索路线的底层逻辑是把计算机擅长的事还给计算机——「搜索、检索、规模化、暴力扫描」[1200–1440s]。

这就是他的「哑锤（dumb hammer）」工具哲学：工具没有智能，锤子只管砸；如果 LLM 要搬运数据、造哈希结构、往数据库里塞东西，你的工具就设计砸了。Agent 负责模糊的智能决策，工具负责确定性地干活，像数据库一样。「我用 Jev 清理桌面？兄弟，一个 shell 脚本就够了。」PII 清理同理：第一道防线是正则这类哑工具，Jev 只是纵深防御的第二层。Sam 给出同构的版本：Unix 哲学、窄工具——「开发票、改发票，一个工具一件事；而不是把 QuickBooks 的钥匙丢给 agent 让它自己看 schema 摸索」。Jeffries 还补了一个自我修正：新型决策模型（他点了 Jev，本站 10 月 8 日专访过其创始人 Diogo Almeida，两场互为印证）让「锤子开始有魔法」——快速小决策（PII 判定、发票红还是蓝）可以内建进工具；他正在攒一个「中脑」架构：本地快速反射的小模型+云端大规划慢模型，与 Sam 此前在播客里讨论的 fast/slow 思维呼应[1440–1680s]。

隐私实践与本地模型的真实瓶颈

具体用什么模型？Jeffries 的答案是分场景路由：工作营销类命名空间里 Claude、转写名 Saul、Astra（疑为 Astra，未确认）随便用；涉及健康、税务等敏感内容则几乎全走 Venice——Erik Voorhees 创办的「关心隐私的 OpenRouter / 模型界的 VPN」，自己跑开源模型、对请求做混捆去标识化、给模型标隐私等级（嘉宾称其年经常性收入已约一亿美元、比 OpenRouter 贵但值得；他催了很多次团队版）。他的隐私模型清单：Kimi K3、GLM-5.3、DeepSeek V4、Flash 4.1。他也坦白自己「被吸进去了」：Codex 天天用、常为自己分享过的信息后怕——「把报税单丢给 GPT 很诱人，我懂」。Nautilo 正在推 BYOK（自带 key）：社区版无法补贴两百万美元算力，用户插自己的 Venice key 即可解锁包括 Seedance 视频在内的全部 Genie 能力[1560–1680s, 2160–2400s]。

本地模型的判断很清醒：Nautilo 暂不支持本地模型，因为还没遇到让他兴奋的——瓶颈不是 GPU 而是内存与量化（他期待苹果下月发布的 512GB 内存新机）；现状是「五帧跑游戏」：重度量化、十 token 每秒、本地视频模型生成十秒片段要三十分钟。但企业侧自建（GB200/300 上起一个 OpenAI 风格端点）几天就能接；真正的规模化版本（容器+vLLM+监控+路由）才是企业要造的东西。他押一个六个月内兑现的路径：某个 27B 级、可本地微调的 Jev 类决策模型成为路由与守门层——在你把数据发给「天上那台受补贴的大算力」之前，本地小模型先把人脸、社保号这类 PII 洗掉[1680–2160s]。

对补贴算力他有一套完整的时间窗策略：当下是「八十年代第一口免费」的毒品经济学、Uber 三美元横穿城市的时代，「音乐椅总有停的一天」；所以趁补贴期用 LLM 把确定性工具造出来——造一次、永远用，而且确定性工具本来就更好。「太多人把不需要 LLM 的任务交给 LLM。」与他强调的共创（co-creativity）同源：他做站点视频时自己录音、Genie 转写成 MD 文件当场可改、试配音、出风格稿、他批脚本再分层合成——「省下几小时，但不是『请给我东西、吐出来啥都当金子』」。他引某位知名 VC 的话收尾：「别再给我发 AI 生成的 deck 了，一股没下过功夫的味道。」[2400–2760s]

「那堵墙」：SaaS 启示录认错与定制软件时代

被问「在今天的環境里造产品是什么体验」，Jeffries 说圈内创业者管它叫 The Wall（他顺势比了个 Pink Floyd）：变现、护城河、注意力、噪声四重困难——工具人人相同、东西都能建出来，于是「被看见」成了主要矛盾；「有时候你看一个东西爆了，完全不明白为什么，你自己试过的十个更好的工具无声无息」。他转述一位朋友的两个新原则：一、我有一个想法时，默认别人昨天就有且昨天已发布；二、不再从商业模式倒推，而是从营销倒推——什么能塞进一条带视频的推文、什么「沙丘遇上恐龙」式一句话能过 lizard brain 那一关。他自己的状态是「极度兴奋+极度疲惫并存」：「都说 AI 要把工作都拿走了，我这辈子从没这么拼命干过。」[2760–3000s]

他当众认了一个错：一年半前他说「SaaS 启示录是扯淡，没人会重写这些应用」——现在承认错了：一批高估值公司正在破产或以低估值九成的跳楼价甩卖（他点了 Miro），而他自己在 Nautilo 里用半周复刻了一块 Miro 白板，Genie 还能直接操控它；前阵子开源的 Photoshop 替代品 Compositor 也被设计师社区迅速围拢。他的推论是「定制软件时代」：微软的老笑话是「你只用了 Word 的 5%，而每个人用的 5% 都不一样」——现在你可以只造你要的那 5%；两三年内（时间线照例模糊）agent 能直接吐出可运行的字节码、随口重构，「那会是一个非常难卖软件的环境」。还剩下的护城河在监管、法律摩擦与软硬件结合处——他举 DeepMind 早年与英国眼科合作的教训（转写 Suleyman，应为 Mustafa Suleyman）：技术能检测眼疾，最终死于官僚体系与《卫报》的隐私报道，「他们如今还跑着那些老年系统」[3000–3211s]。

行业机制与读者启发

  1. 补贴算力 → 建造归零、分发成为唯一瓶颈。机制：免费/受补贴的智能让任何人一周末造出软件 → 产品数量爆炸、注意力稀缺 → 创业策略从「商业模式倒推」翻转为「营销倒推」；SaaS 的命运分化：无监管摩擦的通用工具被定制重写（Miro 式甩卖），有监管/硬件/法律摩擦的反而获得护城河。二阶影响：软件公司的估值逻辑从功能覆盖度迁移到摩擦占有量。观察指标：高估值 SaaS 的甩卖名单、开源替代品的社区成型速度、agent 直吐字节码何时可用。启发：个人与中小团队的机会恰在「有主见的取舍」——大厂通用产品不能做的事（隐私、自主权、人格连续性）就是差异化本身。
  2. 个人 AI 成为世界接口 → 亲密数据入口 → 隐私变成产品维度。机制：接口越贴心，分享的越是最私密的内容（孩子、官司、健康）→ SaaS 形态等于监控经济平方 → 自主权方案（开源+自托管+分布式密钥+时间限定解密委托+命名空间）与隐私路由（Venice，约一亿美元 run rate）崛起。启发：给「发给模型的内容」分级并路由，是当下就能落地的个人/企业实践；考察任何个人 AI 产品，先问密钥在谁手里、记忆能否按命名空间隔离。
  3. 多 Agent 的实用主义边界。同一个底层模型互相开会是角色扮演；有效形态是「一个全知的超 agent + 按需匿名子 agent（牛群）+ 人机同场协作」。上下文永远不够（Mr. Meeseeks），记忆的正解是压缩哈希+检索+衰减而非更大窗口。哑锤哲学给出工具设计的判据：凡 LLM 在搬运数据、就是工具设计失败——这条判据可以直接拿去 review 任何 agent 系统的工具层。
  4. 用补贴造确定性资产。补贴期把 LLM 当「工厂」而非「运行时」：造一次、永远用的确定性工具既省钱又更可靠；等音乐椅停下（Uber 从三美元涨到七十美元），你手里留下的是资产而不是账单。

归属与转写说明

本文事实与引语均出自本集英文转写并逐段核对时间戳；Nautilo 的产品事实（开源发布、Genie、命名空间、E2E 委托解密、Venice 集成、控制 Codex/Hermes/Claude Code、Seedance/MiniMax H3 视频生成）已与 Dan Jeffries 的官方发布帖交叉核对一致。Venice 年经常性收入约一亿美元、Miro 九折甩卖、Muse 泄露家庭住址、高估值 SaaS 破产潮等均为嘉宾或节目口径，未独立核实。转写保留待核项：Cantaros（嘉宾所属机构名，拼写未确认）、Saul 与 Astro/Astra（疑为前沿模型名，未能确认）、quad tag（疑为 Slack 相关功能）、blonde brain barrier（疑为 lizard brain barrier）、OpenAI’s Dot（开场举例，归属以节目原话为准）、「iPhone 56/Mac 3.5 时代」等含糊表述。Jev 相关背景可对照本站 10 月 8 日《Jev 创始人谈机器原生智能》一文，两场观点互为印证。

附：转写疑似错误与校正

以下已确认项已在文末《转写原文（修正版）》附录中机械替换，正文均已采用校正后写法：

  • Nodalo/Notolo/Notelo/Nautilus/Tadano→Nautilo：合计 9 处。产品官方名为 Nautilo（nautilo.ai），转写五种变体误听。
  • Two Shaped Objects→Tool-Shaped Objects：1 处；called Akana→called a kanna：1 处（日本木工刨 kanna）。
  • Mister Me seeks→Mr. Meeseeks：1 处。瑞克和莫蒂角色，喻上下文窗口的短命。
  • JLM53→GLM-5.3、Kimmy K3→Kimi K3、Deepseek V4→DeepSeek V4、Flash41→Flash 4.1：各 1 处。模型官方名。
  • Vans key→Venice key、C dance→Seedance、OpenAI and Topic→OpenAI and Anthropic：各 1 处。
  • window from Kua→window from GUI：1 处。
  • Tuckum→Teknium：1 处。Nous Research 成员。
  • VLM→vLLM：3 处。指推理引擎而非视觉语言模型。
  • meat spades→meatspace、Eric Voorhees→Erik Voorhees、like the H3→like the MiniMax H3、Suleiman→Suleyman（Mustafa Suleyman）、Twimble→TWIML：各 1 处。
  • 保留原文待核：Cantaros（机构名拼写）；Saul、Astro/Astra（疑为模型名）；quad tag；blonde brain barrier（疑为 lizard brain barrier）；「in a not a low」（口语粘连，按上下文指 Nautilo）。

转写原文（修正版）

[0s - 120s] Until recently, the big problem with personal agents was how hard they were to use. Sure, you could set up OpenAI over a weekend, but was your non-technical mom going to use it? Over the past few weeks, though, AI agents have crossed a key accessibility threshold. You no longer need to be a geek to get started. Meta’s Muse, OpenAI’s Dot, and a growing list of others are ready to go right out of the box. So, is it game over for people building agents? Not at all. It’s gotten harder to stand out, but there’s real opportunity in the opinionated design choices that don’t make sense for a general-purpose product, and the unique experiences and workflows those choices make possible. My friend Dan Jeffries has spent the past six months making exactly those choices while building Nautilo, a platform for personal and collaborative AI agents, and he’s developed some pretty strong opinions along the way. For example, a single agent that knows everything about you needs to know when to keep its mouth shut, and he thinks a lot of the elaborate multi-agent org charts we’re seeing are basically cosplay. We get into how he’s approaching persistent memory, coordinating work across people and agents, keeping private context private, and where the model should stop and the tool should take over. Tools are dumb hammers. The hammer that has no intelligence it hammers things. So I believe most of the stuff is like if you are exposing the machinery. In other words, if the LLM has to shuttle the data from here, or make the hash structure, or put something in the database, you screwed up your tool. The agent should supply the fuzzy intelligent decision making, and that the tool should do the work for you. So it should work like a database. When people are like, “I use Jev to clean up my desktop,” I’m like, “Dude, you could have done that with a shell script.” I’m Sam Charrington, and this is the TWIML AI podcast. For over a decade, I’ve been exploring the ideas and innovations shaping the future of AI through conversations like this one that help you understand what’s real, what’s next, and what matters. Let’s jump in.

[120s - 240s] I realized, you know, we’re sort of in the like. If you look at the diffusion of the classic diffusion of innovation curve, I was trying to build products for people in like the early adopter, early majority. We’re still in the two percent, like bleeding edge beginning. Like we’re in the we’re really. You think we’re that early? I do. I really do. It’s like the early, like late eighties, early nineties, where you just accepted that your computer would mess up, right? And you’d have to know how to put it back together, right? You would expect it, and we’re not. We are actually not in the in the iPhone fifty six Mac three five era, where you’re like when we’re just you expect it to work perfectly. You don’t know how it works, and where where if it breaks down, you’re like, what happened? And who do I call? Like how does this? What happens? Right. So we’re definitely not there. And so I I wrote an article on the agent, what I call the agency problem, which is like a lot of times people don’t know what to do with agents, and only the kind of like a small percentage of people do. And I tend to find they they tend to be people who are very high agency, right? They tend to be people who are like I have fifty things I always want to do, like five businesses I want to start, thirty books I wanted to write, three movies I wanted to make, fifty trips I wanted to take. They’ve been in the backlog for forever, and I can’t. And now I’ve got an infinity canvas, and it can do anything. And but and it messes up like it’s it’s like that thing. It’s like the commercial where you become this manager, and it’s like well this employee comes to you and he’s kind of an idiot, and you got to like take the keys and like redo it. And this one comes to you and is like, I I did the job, boss. And you’re like, no, you didn’t. What is this? Like go back. And another one is like, oh, I did it. And you’re like, oh, that’s almost right. Hold on a second. Yeah, that’s good. We’ll just tweak this and this. But you don’t know which of those you’re going to get in the slot machine, right? And you know, and and that and so there’s this weird management era, and then certain things just don’t work. Like I shared this tweet not too long ago that kind of characterized this experience for me. It was like a pie chart of 2026 with a 10 percent sliver of me actually doing.

[240s - 360s] The work and ninety percent sliver of me working on the agents that I’m trying to get to do the work for me. You know, some combination of you know coding, review, iteration. Like, you know, it’s both frustrating, but also it feels super productive in the long term. Like, you know, there used to be a time when you know I I get frustrated when I have to do repetitive tasks that I don’t think are interesting or or valuable, and so I’d spend a really long time trying to you know write the macro in Excel so that I didn’t need to copy all the cells and do all that shit. And now you can you can go down that path like super easily and and really quickly automate a lot of things. But you know those things break all the time, or like you end up finding the corner cases, and like you need to you know go back into this code. Like I agree with your point about kind of this management and delegation era, and how it allows you to kind of expand your plate and do more. Yeah, and I so I had a best friend who was a pretty good tech guy come to me and is like, I want to know what to do with agents. So how do I get started? And I’m like, the question is nonsensical to me as a person who is like in that two percentage like weird high agency semi autistic kind of people, right? Who is like, you know, and I’m like, I don’t understand. What do you mean? How do you get started? Like you talk to it, and then when it messes up, you talk to it again, and you figure out what’s wrong. You critical think and whatever. But there’s like a lot of people are like, well, I just want to. I want the I want the template, right? I want the thing that tells me how to do the thing or whatever. And so I would say it’s just not. It’s not there yet. It’s not in the like you know. It just hasn’t gotten to that phase, and so you spend a lot of time just trying to get the agent do what you want. There was a great essay about it called Tool-Shaped Objects. Okay, and they talked about a favorite, a famous Japanese craftsman wood planer called a kanna. It’s a huge piece of work of art. Takes multiple

[360s - 480s] Multiple craftsmen months to make and set up for you and all this stuff. And when you know it shaves wood with like angelic precision, it’s also economically totally worthless. Okay, it because because a power planer will do it for a fraction of the cost in in in like in a fraction of the time. So like the actual problem in much of the AI today is that the the a lot of it is the setup exists like the con exists so that the setup exists. And when you see a lot of the videos, they’re like, “This is insane! I just rolled out the new thing,” and you’re like, “It’s ninety percent of them setting it up.” And then you’re like, “Okay, great. Now they’re going to do real work with this thing, right?” And then that that part of the video doesn’t happen, right? There’s no there’s no work being done, or it’s like I set up my fifteen what I call the faceless toaster bots, right? I set up my email bot and my and my and then my CEO bot to manage them. I’m like, you’re just LARPing as a CEO. You’re just cosplaying. You’re not actually doing anything, right? I’m like, what are you doing? Well, I’m getting the agents to talk to themselves. But what are they talking about? Why do they need to talk? Isn’t that the same model underneath? Why are they talking about? Like, what’s happening, right? And so I think that there’s a weird disconnect where for high agency people and people who’ve got an insane backlog, the Infinite Canvas is amazing. And for many other folks, they still are looking at it, going like, I don’t know what this does for me, and they’re looking at it with a bit of fear, right? They’re thinking maybe it’ll it’ll take over. It’s not going to take over. It’s not going to take the jobs. The jobs will will change over time. I’ve always said this. I’m like, we’ve already destroyed all the jobs a thousand times in history. We always change. We always adapt. So the last time we spoke was end of 2024. You were working on computer use agents. That’s that was a great conversation. I’ll refer folks back to it. But you know, talk a little bit about you know what you’ve been up to since then, how you’ve evolved, the way you’re thinking about what you’re doing at Cantaros, and in particular.

[480s - 600s] What you’re doing with Nautilo? Yeah, so still trying to solve computer use. I don’t think that it’s solved. I think I’ve seen a lot of tweets otherwise that promise me it’s solved, and it’s not. Okay, like Jib is still not a visual model. Jib doesn’t always make great decisions, so that means you’ve got to translate the window from GUI into like a non-visual sense. Like, so I’ve been trying to do a lot of stuff there to help crack it. You’ve seen a lot of kind pre-baked workflows that are not actually, you know, correct that are not generalizable, and you still have the round trip to the cloud of like if you’re going to take screenshots. So where we were at was trying to micro train agents to do one thing really well, and but the problem is it didn’t work too well, right? It kind of worked, but it didn’t really work the way we wanted with Agent Tutor, and so we were trying to say I’ll be able to kind of just train it to do this workflow, but it was just it’s just slow. Like now the models are much better at screenshotting and understanding coordinates and all that. They’re all trained on it, but you still you don’t have a micro model that just lives on your desktop that’s Jib fast. That said, models like Astro and Claude are are pretty good if you don’t need to move at human speed for software testing for doing DevOps, ClickOps style work for you. Whereas in the past, as a Dev as a system, I’d have to go in and like figure it out and type the commands myself. Like a lot of times now, you can ask it and watch what it’s doing, and that’s pretty cool. So I think there’s been a lot of progress there, but it’s still not fully solved, right? So then Nautilo, what we pivoted to is we’ve been burning on the concept of personalized AI, right? I think a lot of people are burning on it, but I think we’re thinking about it very differently. So okay, Nautilo, I think my my friend joked is 30 products in one, not one, but but but the but the the one liner is it’s her, but for teams or families, right? In other words, her as in the movie, as in the

[600s - 720s] Movie her right, the flip phone version of her. Okay, not not the full OS version because you can’t build that yet. Not the thing that can plug into anything that has a perfect personality. But what we did is we built we built incredible personality, customizability, radical customizability into what we call our genies, our agents, and they act like secret keepers. We call like they act like secretaries in the old form of the words “secretum,” which you know, which means in the Latin, which means secret keeper, and they were the the keepers of powerful people and their effects, right? And now everybody can have one. And we we wanted them. We had to re we had to rebuild kind of group chat. And we have a magical router in there that’ll allow like the agent to just wake up by just talking naturally, and and you’ll be in there with people and with your agents, and that you can just chat naturally. And they all seem to have their own personality. They keep their own memory. They don’t really seem to break that. We have an amazing memory system. You know, you can’t really do this in the kind we’ve at. We’ve been asked a lot of time, like, well, why didn’t you just plug in every messenger in the world? Well, you can’t just magically route in there, and you’re also violating the terms of service. They don’t really want you there. Like those things were based on they their terms of service were built on when you know bots were a spam, right, and to be kept out. So there, even if you go get a business license from, are we talking about like Slack for example? Well, yes. Well, so so Slack, you can put them in there now with like you know with quad tag and things like that, but it tends to be a singleton AI and not a diverse you know AI, right? That has like a lot of different personality quirks, and like you have to add it every time, like at, and it wakes up and those kinds of things, right? And so I felt you know we needed to like reconstruct that, and we also needed to we needed to give people self sovereign AI where you control the backend, the front end, the mobile clients completely as open source, and whether I win or not, I think this is the way of the future. I think we have to melt down all the software and recraft it, and I do think it has to be self sovereign because as Francois Chollet.

[720s - 840s] He said, “AI is going to become our interface to the world.” Like I say to people, and they think that it’s crazy, but you won’t even know how to open an application on your phone in a decade, right? Like you’ll just talk to it, like in the movie, and it’ll do things for you. And the problem is that that’s another SaaS application, and all the personalized AIs right now, and we’re up against a wall of free compute and infinite budgets, and you know, and and so it’s very hard to break through the noise. And people don’t fully understand what it is that, like, because it’s going to be our interface to the world, it’s going to be much more intimate what you share with it, okay? And we feel that if that is a SaaS application, okay, that it is surveillance economy squared. Like, if you think that what happened, like, with social media is bad to society, it’s going to be infinitely worse when you’re sharing your kid’s problems at school, the mall you’re worried about, you talk to it about a legal problem, and it shows up in court. These kinds of things, right? And and we, the only way to do it is it’s not going to be passed by law, right? Is you’ve got to go back to kind of the cypherpunk days. So we’re building in a spin on end-to-end encryption into it, where the back end is is distributed keys, and unlike end-to-end encryption where it’s just humans, a genie needs to read stuff in plain text. So it gets a time-based delegation to wake up and decrypt an asset, decrypt a chat, decrypt a memory, etc., and talk to you, and then it it stops, right? And so we think that these kinds of things, you have to have a cypherpunk solution of self-sovereign AI. We don’t currently support local models, but we we will. It’s it’s pretty easy, and as they get better, we support things like Venice AI, which is almost like a VPN for AIs, right? And we rate the models by privacy. You’ve mentioned a bunch of different things. Let’s yeah, zoom in on that and zoom in. I guess the first one I want to start on is multi-agent and your experiences with multi-agent. I’ve gone back and forth on this, both through you know conversations and interviews.

[840s - 960s] My own experience, I get the general idea that multi-agent makes sense from a, you know, basically due to limitations of the models, like keeping context about a particular domain constrained within an agent that’s responsible for that domain. But at the same time, my experience, I do a lot with Hermes. I have it dealing with a lot of different domains related to mostly business, less personal, and it just does it does just fine, like pulling the skills it needs for a particular problem. But a single agent, and so I kind of want to hear you talk a little bit about like why should people even care about multi-agent? You know, where have you found that that’s really useful, and like what’s required to make it you know useful? Well, first of all, I love the Hermes team. I think they’re fantastic, and these are doing incredible work. I love Hermes. I love everybody on them. I follow them. They follow me on Twitter. You know, a lot of the folks there, Teknium and Shannon Sands, their their aliases online. There, I think they’re just incredible developers. I think they have an incredible philosophy. I love them. I believe we actually have the same philosophy. I believe in a super powerful genie that is, or a super powerful super agent that knows everything. I, it’s a pets versus cattle thing. Okay, the agent can spin up the cattle, but I don’t want to know the names of the cattle. Okay, like they can spin up sub agents and do things, but I don’t want what I call the faceless toaster bot. I don’t need the email bot. I don’t need the the the the tax bot or whatever, and then I make the CEO bot to manage them. Like, yeah, yeah, yeah, yeah, right. So, so that’s not what I mean by multi-agent. I mean everybody has an agent radically customized to them. They can do anything. We have things it can do video editing. It’s we have co-creative apps built right into it where you can you can take over and trim it. You don’t have to do what I call prompt and pray, where it’s like it makes the video and then you’re like, oh no, like I need to cut out this two seconds. Please cut it. Oh no, iteration fifty six. You can cut it yourself, or you take over and write the the.

[960s - 1080s] The MD file or the Word file or the Excel file right in there, right? Or they take over the terminal. I have a Let Genie control it, whatever. So what I mean multi-agent is humans and agents on an even playing field in a conversation. Now there’s a lot of work that you have to do because the agents are not really used to talking to each other. So you have to build a bunch of little classifiers and backend things that route and tell them to like you know look things up where what memories are shared, etc. Right? And we also include things like name-spacing. So this is a super important thing. You see things on the news where like Muse basically gave someone’s home address and things like that, and that’s because it knows everything and there’s no name-spacing. We built name-spacing into the core so that when you talk about something with your person, your agent in a personal chat one-on-one, which you’re always talking to the same super agent that you’ve personalized, you gave it a name, it’s a it picture you picked, it’s Eleven Labs voice from thirty thousand things, right? You customized it, and so all these kinds of things. It keeps its memory radically, just like in Hermes. I think they did a great job with memory. We did a great job with memory. I’ve never seen my agent kind of break character, if you will, right? And and then you talk to that specific agent about things, and they can do all those things for you. If it needs a sub-agent or whatever, it does that. So I don’t believe in kind of the LARPing approach, the cattle approach of like you know I need to thirty agents to talk about it. This is my marketing agent. This is my. You don’t need it. The same agent. Yeah, yeah, yeah. The same agent underneath it. But so like, but you do have a marketing channel where you and your team are talking about your go-to-market strategy. And here’s an Ask Me Anything doc that we wrote for you know X. Or here’s our initial tweet. Genie, can you check out you know Genie one or Genie two or Genie three? Can you know which they have their personal names, right? Can you know can you go and check this thing out and make an edit on it? Or then your your team member reads it and goes, I don’t really like that line. I’m going to take over and write a little bit. But that’s a convenience for the people, not because the agent you know.

[1080s - 1200s] Needs to be separated into these different contexts. Well, well, it’s it is a convenience for the people because you end up doing what I call like like meatspace like delivery or like pigeon delivery, right? So, like, imagine like we could we could have an agent where let’s say this happens all the time. You you say like you’re going on a trip next summer or whatever, right? And you go do you do about you have your agent go do a bunch of research on it, and then like you now become like the interpreter for your wife, like copy and paste it somewhere, and you and she’s like, well, what what about this part? And you’re like, well, you haven’t fully read that part, right? So now you got to go read it and be the meat space interpreter, right? So what I have, what we built, how do you know what’s going on in my life, Dan? Yeah, because it’s happening in everybody’s life right now. Okay, so so my wife is not not reading that long ChatGPT conversation. Hell no, she is. No, the first thing I do is okay. Now give my wife the TLDR thing. So the thing is, I can now send. I can take the agent. She made the report. I could take it and say, take that memory and that report. Go talk to my wife and get her opinion and come back to me, and it’ll come back to me with what she was saying about it. Like I think we should look at this, this, and this. So you, the agent has the context. Let it be the the agent context for that thing, right? And then it can always search history. It can always go through the history and go. You go. Who was I talking with about this and that one agent you were talking to was like, oh, we were talking about this this stock trader. This person was in charge of this or whatever, and it’ll look on those things and understand. And it’s a it’s a it’s a context keeper and a secret keeper. So it’s not just convenience. We’re acting like meat puppets or pigeon carrier delivery between people when we don’t have full context. So to me, that doesn’t make any sense. So in talking through kind of multi-agent and some of those usage scenarios, you describe what I heard as like a lot of context that the agent needs to be able to manage and manipulate and

[1200s - 1320s] Are you finding that modern models, you know, have large enough context windows that, like, you can be kind of that—that’s less of an issue for you as someone building the system, or are you doing like, like, super surgical applying super surgical heuristics to to context in order to to make it all work? Like, you know, how are you approaching kind of dealing with the context limitations inherent with models? There will never be enough context for me. They’re all Mr. Meeseeks. Okay, so the way that we’ve solved it is, I think, a number of memory breakthroughs. We’re probably going to do a paper on it where we’ve got like compressed hashes, so we can for that are expandable, like to like get a little bit more context, a little bit more context, a little bit more context. We’ve we’ve done the old thing again of what computers do well, which is search and retrieval and scale and brute force. And we’ve given them a tremendous retrieval system. So the first thing you’ll always see the genies doing is looking stuff up, looking through the conversation history, looking through like memories that they have. They have ton and they can keep a ton of memories because they have the hashes put in without it trying to stuff every full memory in that they can then go retrieve more of that memory. And the memories decay over time. They’ll decay the memories. What is the hash? The hash basically is is a compressed version of us to say like it’s almost like a lookup list of like meaning. So you’ve got a like the agent at some point has to be reading plain text. You’ve got this right. So yeah, you have so you have yeah. You have a little to other things. Correct. You have a little bit of metadata and a hash that’s like a clue that it can go. Okay, I think these hashes are probably the things I need to go check out a bit more. Let me expand that and get like more than a snippet right of that of that, and then another snippet. And go nope, these three I don’t need. This one is probably it. Let me expand that one fully and read it. Okay, and then kind of come back before I open my mouth. That has been tremendously effective in terms of like I’m often shocked with like how well it seems to remember things.

[1320s - 1440s] And like I said, I’ve never seen our genies break character, which I’ve seen, and we’re also sometimes shocked with as they build up their own personalized context over time, either in group chats or about you. Even when they’re using the same models, they tend to have a diversity that is sort of shocking to us, right? And and we can’t fully explain it actually, even at this point, right? So are the the agents like building this kind of hash memory table structure thing constantly? Is there deterministically done? Got it. So that’s the harness. I have a philosophy. Determinist. Correct. Correct. I have a I have a philosophy of tool design that I call, and it’s it’s got a little fuzzy lately, but and I’ll explain that. But it’s what I call the dumb hammer design, okay? Or the magical hammer design, okay? And what I used to see in the old agents is the agent builds magical capability into the code, like the code makes an LLM or human style decision in code, right? Like a fuzzy understanding is determined that, and so I would see old agents do this, and I called it the magical hammer problem. And so I have a rule in there where I’m like, tools are dumb hammers. The hammer has no intelligence. It hammers things, right? And so I believe most of the stuff is like if you are exposing the machinery, in other words, if the LLM has to shuttle the data from here or make the hash structure or put something in the database, you screwed up your tool. The the the the the agent should supply the fuzzy intelligent decision making, and that the tool should do the work for you, so it should work like a database. When people are like, I use Jev to clean up my desktop. I’m like, dude, you could have done that with a shell script, okay? You didn’t. You didn’t. That’s you know. I’m sorry. You know what I’m saying? Or they’re like, I’m going to use Jev to like, you know, to strip PII. That’s secondary defense in depth. The first thing you should have done is used a dumb hammer, right? You should have used like a like you know.

[1440s - 1560s] You know, a search for like certain patterns, right? Regex or something with regexes, right? You know, and the millions of tools that are out there, and then on top of that, blur the faces and things like that that are hard, right? I’ve kind of evolved a very similar philosophy about tools, and it’s not so much about dumb versus smart, but it’s about like narrow. Like I’ve had a lot of success with okay, this one tool. It’s a little thing. It does this one thing. Unix philosophy. And the you know the the agent if it needs to do this one thing it doesn’t need to think about anything it just needs to call this tool with the right parameters and the thing gets done like that’s right create an invoice edit invoice do this do that like no I’m not just I’m not giving you the agent my keys to QuickBooks and have you go figure it out by looking at the schema no these are the things you need to do do these little things and we’ll be a lot better and it’s worked out really well so far I believe in the Unix philosophy and the Unix philosophy was always that like every tool does one thing well right like it and and the Windows philosophy back in the day was I have a monolith that does 80 things poorly right but like the Unix philosophy was always like grep greps you know less is more more is more right you know what I mean like the ps gives me the processes and filters them that’s it okay and grep does it even better the filter right and awk and sed do it even better right in different ways so those are the things where like awk and sed should do a lot of the heavy lifting for you and then the agents to apply the fuzzy context now I will say that with the new decision style models and I do I kind of ragged on Jeb a little bit but I think it’s freaking awesome and I think the I’m really looking forward to the decision visual versions of it I think it’s the kind of thing where you make a small model now and it’s going to be the type of thing where local people you know local models can start to excel again because they can fine tune on these things and they’re going to be visual I think I think these are

[1560s - 1680s] Really cool, where they can make lots of decisions in a short period of time. Now that does kind of give you a magical hammer, where you can kind of build in like, which way should I go? Right? Should I go left or right? Should I like jump up and down? Should I like classify this as PII or not? Should I make the invoice red or blue? Right? You know these kinds of things. So I do think you’re starting to get to like a fast and slow level, like thinking, you know, of intuition, which is cool. And I think we’re gonna. I think what we’re missing, and I’ve been trying to build into my latest version of, in my endless search for like perfect computer control, is like the mid brain, right? Which is like a mid level flash style model that can make like semi complicated decisions quickly, right? And so it’s like lightning fast decisions, semi complicated, and then like the big planning slowing model, right? Right? And kind of like separate that into like fast reflex, muscle twitch, and and slow, and kind of trying to find the balance there with it. And I think, I think I might crack it. But you know, this is I thought I might crack it two years ago too, and four years before that. So. You mentioned Venice in there. Yeah. Tell us about Venice. What is that? No, Erik Voorhees started it. He’s a he’s an old school sort of libertarian. He came out of the kind of that kind of thinking of you know the cypherpunks and things like that. So he they’ve moved to I think the last time I saw something like a hundred million dollar run rate. Those are ones where they run a lot of the models themselves, the open source models, or they and they rate the models like so they will give access to things like OpenAI and Anthropic, but they’ll try to like pool them or they’ll say like you know those or bundle those requests so that they’re not sort of like you know. De-identifiable or something. Yeah, de-identifiable, but and they’ll rate the model kind of based on it. And then they have kind of their own models that they host that say like you know we’re not going to keep all of the the stuff on this. And I think. So is it? Can you could you characterize it very loosely?

[1680s - 1800s] As open router that cares about privacy, yeah, it’s open router that cares about privacy. I’d say it’s a VPN edge for models, right? And you know, I think I think they’re doing God’s work. I think they’re doing amazing stuff with it. I have all the respect in the world for Eric and his team. Although Eric, I’d like to see some team-based version of it, not just individual subs. Please, you stop. I don’t know how many times I have to send an email or tweet at you out about, but please, please, team enterprise version, okay? But otherwise, look, I mean, I think God bless them. They’re doing amazing stuff, and I think those. I’d like to see more of that kind of stuff, more diversity. And I’m just, as you know, my my roots in open source go back a long way. I’m constantly fighting with it. I can’t believe that we’re actually fighting the battle again. I thought that this was settled, like started decisus. I thought it was done, but apparently not. But Nautilo does not today support local models. Not yet, and primarily because I just haven’t found any models that I was super excited about locally. I think, I hope that that’s going to change with the new M5, M6, 512 megabytes, 512 gigabytes of RAM units coming from Apple next month. I think that the the real problem, the real bottleneck, is not the GPUs at this point. The real bottleneck is the memory, and and how much you have to quantize the models. That said, we’re compressing more and more intelligence into smaller things. It would not be hard for us to support a module already, probably in a day or two, to say like if you’re running this on your own hardware on big GPUs, you know, GB 200s, 300s, or whatever in your own thing as an enterprise, that you know we stand up an OpenAI style endpoint or whatever that we probably do that in a couple of days, right? So, I I would like to get there, but I I built we built it we built it to be enterprise ready, where like you know we have a CLI for deploying it with drivers.

[1800s - 1920s] To like railway and to you know droplets, you know would be easy to add GCP, you know would be you know, but we didn’t build all of the super scaling out kind of stuff into it yet, right? Or the host, but but that’s where that’s where it will go. And again, whether I win or not, I think that enterprises are going to build this because the stuff, the open weights models are going to are going to get to the point where you you hit a an S every infinite curve is a exponential curve is an S curve, and at some point you just don’t like it’s solving the Poincare conjecture and doing like the DNA you know work, but you don’t need that like Merck needs that, you know like like maybe the quant you know the quant traders in like you know Black Blackstone need that, but I don’t need I needed to like once it’s computer you solved it, it can make a funny video or a presentation or whatever, and and talking about life about our travel plans at some point like you’ve got good enough, and I I think we hit that in in the next couple of years. You think it’s possible that on the local side, like getting that up and running looks very different from you know just a single endpoint in the cloud, meaning you say you can get it up and running a couple days, but like the reality may be that like you really have like a suite of models and you have to have much more sophisticated routing because you’re trying to like optimize to. Yeah, so yeah, look, I I believe that the enterprise version of it is not is not a couple days. Let me let me be clear, right? Yeah, yeah. Supporting supporting vLLM and a Docker container or whatever or vLLM with with your local model for you, you know your or your your HEB, you know your your eight way cluster or whatever is feasible in a couple of days, okay? But like the true version is going to be a scale out version with routing where you’ve got like you know tons of containers and in vLLM and and and monitoring and DevOps system. So I I believe that’s where enterprises are going to be. Yeah.

[1920s - 2040s] Yeah, and not too much thinking about it from less from the perspective of like enterprises do complicated stuff, but just that like you know we we can think about intelligence as in the cloud as like you know this one surface that we push stuff to, but you know maybe intelligence and open weight local models is going to be lumpier and some models are good at some things and other models are good at other things and like in order to have a similar experience locally, you know sooner rather than you know later, like we have to be more sophisticated about what model gets what task. Yeah, so I I think there’s actually a short term potential path branching pathway of the future in my Monte Carlo simulation where Jevs models because they’re relatively small, we’re seeing a lot of people like training them, you know, in a couple of weeks, someone will find the right recipe for doing that, a 27b model or a small model that that can run locally on your computer. I think that actually becomes a cool routing layer. Are you talking about using Jev and these 27b models or training a 27b model like a Jev? Yeah, a Jev like model running locally could be incredibly valuable to say, you know what, I need this to control. It’s visual and needs to control my my. I don’t need to go off to the cloud to like click and open Word and and put the thing in there. I just need that open, and then the big model in the cloud can write the machine consciousness poem in there, right? And I think also it can be the kind of thing where you’ve got a lot of potential with security. In other words, you have the ability, like I said, to use the dumb hammer heuristics of like I’ve before I send this off to the big friendly like subsidized compute in the sky, I should maybe scrub this right for faces and P I and all this, and that you could have a local Jev model going, hey, you your classifier missed some stuff, but I caught a social security number there and I’ve blanked it out. I think that can be very cool. I think that can I think that happens.

[2040s - 2160s] In the next six months, and I think that’s going to be very cool. So that that would probably be the point where I would start to say this is valuable, even before I have the the five twelve gigabyte machine, super fast or whatever. It’s just got to be relatively quick. So a lot of times I always look at it, and there’s a lot of people I respect working on the local models, right? And and I and I and I’m like, you know, I’m like right there with like you know like the self sovereign thing. I believe it. But a lot of times I look at it, I’m like, well, it’s definitely a tool shaped object to set up. It’s challenging, and it’s also you know a lot of times they’re like, ah bam, I got it running. It’s you know it’s super quantized and it’s like ten tokens a second, and I’m like, okay, that’s not quite where I want it to be yet, right? You know what I mean? So it’s kind of like it’s sort of like yeah, I got this GPU and I can play the game at five frames per second. I’m like, not quite what I’m looking for yet, right? Or like I was very excited to see like the the video model like the MiniMax H3 or whatever and things like that was like, oh man, this can run on local hardware. This is awesome. And it was like the average things were like thirty minutes for generation of like a ten second video. You know, and I’m like, okay, that’s that’s you know that’s that’s not going to work for me. Like, but I do think that that gets better. But it’s really a like Steve Jobs said, the software is just all the stuff you didn’t have time to get into hardware, right? So like as this hardware gets better, right, then it becomes it becomes an easier problem. And so I think I do think hardware just moves slower and it takes a little bit of time and the and and a lot of kind of tricks to kind of get us there. So what models are you you using? Given your given the degree to which you care about privacy and self sovereignty, when you are using your own Nautilo, like what models are you using? How are you thinking about kind of ensuring that you have the degree of privacy you want with the constraints of needing to use cloud based frontier models? Yeah, so I think it depends on some.

[2160s - 2280s] Sometimes the chat that I’m into, remember we have namespaces. So if I’m working on, if I’m working on the marketing stuff, that I have no problem using, you know, Claude or Saul or Astro. Okay, so I just want to be clear about that. I love those models, and I, and I, I’m not, I’m not an open, you know, or privacy absolutist. I believe you have diversity. I believe you have a need for privacy, and you have a need for open weights and open source, and you have a need for commercial stuff, and they can all and should all coexist. That’s fine, you know. But when I’m using privacy ones, I tend to use the, I tend to use Venice, and I tend to use, I tend to use Kimi K3. I tend, I tend to use GLM-5.3, DeepSeek V4, Flash 4.1. All on Venice or via. Almost all on Venice, right? Yeah, almost all on Venice. Those tend to be the ones that, like, if I’m asking about a health issue or something like that, I tend to, I tend to, I tend to look there first. And I think I would encourage people to do the same. I would encourage people to think about the kinds of information that they’re sharing with it. It’s really tempting to not care and to just say like, “Here’s my tax return, GPT.” Like, I get it. Like, GPTs look. There’s no shade, man. Like, I mean, you know, Codex is incredible. I use it all the time. It is. It is incredible. And when I think about it, I’m horrified at how I’ve gotten sucked in to sharing some of the information that I share. Like, yeah. And it’s very reluctantly. It’s very consciously. It’s probably not nearly as much as people who aren’t thinking about it are willing to share. You know, but there are convene and there are you know conveniences that I give up because I’m not willing to share certain information. I agree with the the hesitation to share, and also interesting to me to learn about Venice. I’m going to check that out. Yeah, they’re great, and they plug into almost any harness that you want. They’ve got a pretty good. You you pay a little bit of a premium for it, which, which I think is fair. They’re more expensive than OpenRouter, but I think.

[2280s - 2400s] They’re not sending you off to Kimmy in China. They’re they’re running it on their own hardware, right? And so that’s the beauty of OpenWeights, right? And I think that’s really important in their policies, the way that they think about things. So that can go into any harness. They have their own subscriptions now. Maybe they don’t have as much subsidized compute as everyone else, but you know, there it’s it’s worth giving them a subscription, you know, for certain types of things. So actually, one of the things we’re doing now is we’re we’re we’re looking at some kind of like you know some givebacks in inference to defray costs, right? And and we’re looking at like originally we designed Nautilo to be like you have your keys on the server by your sysadmin, and everybody uses those, right? But now we’re we’re bringing bring your own keys. And this was someone in the community was like, well, I just want to plug in my keys because I had the community, and I’m like, well, I’m not opening. I can’t afford two million in compute to give away on the community so everybody can use the genie free for eternity. So I was like, okay, people just come in and they’re using it like Discord and chatting with me, but they’re not getting the full power of it. But if you bring your own keys and you go, hey, I’m plugging in this, then we can plug in and say, here’s you know here’s a Venice key and and it gives you access to Seedance and to and to music generation into um into the LLMs and to Kimmy and all these kinds of things, and you get the full power of playing around with the genies and like in in wonderful and it becomes almost like a new use case, almost like a Discord with uh you know with with genies as first class citizens. So I think that’s that’s we’re working on that right now. We hope to have that out next month. I think it’s wonderful, but you really you got to think about you have to think about these things a little because they’re going to give you you know it’s like the old it’s like the old like I don’t think they would tell you during the 80s, right? I think you’re old enough for this too. Where like yeah, they’re going to give you the first hit free. That never happened, by the way. Those liars, those those liars, you know. But um but like yeah but like you know and then you’re going to get sucked in, you know. Well, you are.

[2400s - 2520s] Like yeah, everybody’s getting sucked into the subsidized compute. But at one point, at one point, and me too, right? At one point, the you know, and we and we built actually in Nautilo, we built the ability to control third-party harnesses. You can control Codex and and Claude and Hermes, whatever, right, from inside, and it acts like you know, which is great because people want to use it. But at some point, the musical chairs stop, and and you don’t know when that’s going to be, right? But it’s like the early again, it’s like the early days of of Uber and Lyft, where like it’s like it’s like I’ve used that analogy as well. Yeah, three days, three dollars to get across the city, and now it’s seventy dollars to get across the city, right? Part of my answer to that has been to use the subsidized compute to create deterministic tools to do the things that you want to do. Like, and I don’t think enough people are doing that. Like, it’s easy to say, okay, here’s my task. I’m giving my task to the LLM. The LLM kind of does a good job at the task. I’m like, but your task doesn’t really need the LLM. Like, you can build like you know workflow tools using the LLM. Build them once, you have them forever, and you know they’re actually better because like it’s deterministic. But in the places that it can be deterministic, I don’t think enough people are doing that. And and people aren’t always doing all the work. Like you know, like I still think I still love co-creativity. So like I’ve got one of the videos on the site where I show how I made the videos for the site. So like I wanted to be careful of too much PII. So I made a whole like mock company with different people, characters, and genie characters, so that I could use that rather than like all of our personal stuff. But like when I record the video, I’m recording it with my voice. So then I have a whole video of the making of the video where I’m talking to my genie. I’m like, use my voice, use the video, go, give me the transcript, and then it shows me an MD file in there, and then I can edit the MD file directly where it messed up. And I go, okay, now go, you know, go audition some voices for me, and it’ll audition different.

[2520s - 2640s] Voices for like what I want, you know, like a you know a a wizened old you know you know dude with bald head, right, you know, and um you know and like with you know with a deep voice, you know, or whatever, and and then and then we get voices that we like, and she would just go take those parts and give me back the transcript, then I would take it into the video editor and cut it in, right, and like and overlay like split the audio video and put that in where my voice was, and then recut the video, and then then I might give it back to something like Astra or whatever, and say, okay, like first, let’s do a style design, do some mockups of like what this looks like. Don’t just make it, like let me see. I don’t like this, and I understand kerning, and I understand, you know, I understand taste, and so I could be like, no, that’s not the look that I’m looking for. So that’s the one. Okay, now go in and like like give me a script. Okay, I don’t like that script. I’m going to rewrite it. Okay, that come back. Now I approve the script. Okay, now go ahead, layered in. Now that saves me hours of time, but it’s not here. Please give me the thing, and whatever poops out of it is, you know, is golden. That’s that’s a thing where like I I think I saw a famous VC yesterday say, you know, just stop sending me your AI generated decks. All they they reek of no effort, and I think that’s the kind of thing where like if you don’t put any effort into it, I understand why people get upset about that, right? When they’re like, you just pass off some essay as your own or some deck as your own where you clearly have not stepped back and done the critical thinking or re edited or been a part of the process yourself, and and there you know there are people who are like full on Luddites where like if you don’t do it 100 like if you don’t thresh if you don’t if you don’t thresh the wheat yourself, you’re not a real man. I’m like, all right, listen, I’m I’m sorry, I’m using the lawnmower, okay? I’m using the, I’m I’m I’m using I know there’s there’s like I can hand saw it or I could use this, but I’m going to use the power saw. I’m sorry, okay? Like yeah yeah, but but um but I do think co creativity, putting effort in a thing is just too underrated, and I think a.

[2640s - 2760s] A lot of the times, like you know, Marie and I joke that sometimes, you know, we build these co-creative tools because that’s we love. And a lot of times, people are like, “I just wanted to do it for me,” and you are like, “Okay, like I understand that, right?” And like a lot of times, there are things where I am just happy that it does pieces of it for me or parts of it. But I still think if you are doing something that matters to you, you are pitching something, you are starting a business, you are rolling out your marketing. If you are not even looking at it, what the hell are you doing? What the hell are you doing? One of those co-creative tools is like a co-writing tool that you were working on prior to Nautilo that you’ve kind of integrated in. And I remember reaching out to you and like just wondering, like, why, why isn’t this more common? Like, it’s you know, I think OpenAI just at Dev Days yesterday announced like a a co-writing tool. If I interpreted it correctly, I haven’t looked into it very deeply. But like you know, all AI should come out. You know, should you know, give you results that you can manipulate. Why do I have to you know every time copy it? You know, put it back into the paste box and say, okay, you know, let’s advance this. Let’s you know do this, do that. Like that seems to be such a natural paradigm, and it’s not nearly as common as you know I would have expected it to be by now. Like I found in the earlier version of the of our of our software that most people like just write this for me, man. Like in their prompts, and we’re like, oh God, right? Like so it was just sort of like, why won’t you write it for me? You know, and and so it was just sort of like, okay, like apparently we’re just selling to people who don’t exist yet. Yeah, one more thing I wanted to dig into that you’ve touched on, I think, a couple times, but just to kind of pull it out, like you know, we’ve talked about you know, kind of the age of free compute. You have all these these resources, while you can take advantage of some of that free compute, but you are also competing against it in some way.

[2760s - 2880s] Is like, I think what I want you to kind of spend some time on is just like your experience, like building in this environment, like trying to create something for people, like in today’s world. Like, what is that? Like, you know, what are the the key challenges that you deal with, and like, how are you thinking about overcoming them? No, so me and and several other entrepreneurs that I know are talking about this all the time, and we call it the wall. The wall. The wall. The wall. Pink Floyd. Pink Floyd’s the wall, right? Like the like another giant wall, and that is like you have the problem of monetization, you have the problem of moat, you have the problem of attention and noise, you have the problem of like everyone with the same tools, and so while it’s simultaneously easy to get something built and get something out there, it’s very hard to you know be seen. And sometimes you look at stuff and you’re like, why is that like going viral? And you really have no idea. Like you’re like, there were 10 better tools that you personally tried that just didn’t go anywhere. And like, look, perception is reality. And so I’m, you know, I was having a conversation with a friend of mine, and he was saying he was always he’s got a couple of principles now, and and I agree with these four. He said, if I have an idea, I assume that somebody else had that idea yesterday and is putting it out yesterday. Okay. The the second thing is he said I used to say like what’s the business model and how do I like you know how do I who’s your target audience what’s you know what are you going to do right and work backwards from there and now he’s like I’m working backwards from the marketing like what can I fit into the tweet with the video with the one like with the Dune meets dinosaurs Hollywood elevator pitch that passes the blonde brain barrier and goes viral.

[2880s - 3000s] Right and and so there there’s definitely like there’s an adaptability that I have not fully mastered, right? Like I feel like it’s sort of like the old kung fu movies where you know Kung Fu Panda comes in and just you know gets his butt kicked by the the the spinning gyre and the the cup of fire and the balancing. You’re like whoo, right? Trying to navigate your way through this. I at the same time it’s like I feel like the the Omari guy, you know, who’s like enthusiastic, like I’ve got the greatest tools of all time, and like I why wouldn’t you be excited about this? It can do anything. It’s a it’s a magical genie that you ask it to do stuff and it does that. And so I simultaneously feel this like super excitement and a ton of exhaustion. I’m like, and people said it’s taking all the jobs. I’ve never worked harder in my life. I don’t know what they’re talking about, right? So I you know I’m waiting for all the jobs so I can chill out more. But that’s not happening. I think it is an incredibly challenging environment. I think we’re going to see a lot of folks crash out, and I think it’s very hard to be heard through the noise, even when you have something that’s really really cool. And and then oftentimes sometimes a product comes out and people are excited about it, and then it just disappears. So I think I think in the past you could build Miro, and now you see a lot of SaaS companies. I was actually I I always like to say my predictions that were wrong, and I think about a year and a half ago I was saying the SaaS apocalypse is nonsense. Nobody’s going to rebuild half of these applications. I think that I was I was wrong about that. I think a lot of companies will not do it, but you are starting to see a number of companies that were worth a huge amount of money go under and be sold in fire sales at 90% below value. Miro is just one of them and others. I built a Miro board in a not a low in in like half a week. That’s

[3000s - 3120s] It’s pretty awesome, right? That my genie can control too, you know. So the and I and that you saw someone had the compositor out the other day, and people jumped on and it was like a Photoshop alternative, you know, that was pretty slick. This was a designer. This wasn’t a person who was like, I don’t know anything about Photoshop and I’m buying it. It was like, I mean, I looked at the repo and there were a few things like in the in the readme about how to how to sign it and check it in. I’m like, take that out of there, right? You know, so it helps to have dev devs in there, right? Um, so like, but again, a community project forms around that, and you know, yeah, it does. It have every brush and whatever, but like, what I think is you you have this bespoke era of software coming, where like there. So someone at Microsoft told me the joke in Microsoft used to be, well, you you you only use five percent of Word. Well, everybody uses a different five percent. But now I think the the answer is like you can build the five percent of the things that you really need, and I think probably in two or three years, maybe it’s a little longer. The the timelines are always fuzzy. You are going to have a point where the agents can just spit out byte code of that basically does what you want. If you don’t like it, you say go test it and change it and refactor in this way on the fly. And and so I I do think that that’s a very hard environment to sell software in. Now there’s still going to be software that’s incredibly challenging to build, and and those are going to be in areas that are ironically super regulated, healthcare, right? Like you know, in the infinite infinite mind about the Demis and and the early days of Deep Mind, they talk about um uh Suleyman, you know, doing a bunch of stuff in healthcare with like the UK where they had like worked with like people in opticians and they had a software that was able to detect and all of it died because it it ran up against bureaucracy. It ran up against pieces in the Guardian that were like the the data privacy, and so they’re still running these geriatric systems. So I think there’s a lot of things where.

[3120s - 3211s] People don’t understand the friction, you know, and and where things are going to be. There’s going to be barriers in there, where where they might create a legal mode or there might be a friction mode or whatever those kinds of things, or that where there’s a combination of hardware software that’s really hard to solve. But a lot of other stuff, man, is super hard to have a mode in the competition. And I don’t know where somebody gets a legitimate exit in three or four years that other other people haven’t been able to build. Well, keep on fighting a good fight. Power to the sovereign AI people, and I’m going to keep fighting, and and I hope everyone else will. I hope everyone else will wake up. But don’t say I didn’t warn you. Awesome. Well, Dan, it’s always great catching up. Pleasure. And you know, hearing a bit about what you’ve been working on. You know, as is always the case, the conversation meanders a bit, but always a good one. Well, I’m sure I’m sure you’ll you’ll chop it down and and up. I’m I’m like a squirrel, you know. I got I got a fifty things I’m burning on at once, but I appreciate you indulging me and reaching out to me as always, sir. And it’s good to see you still fighting the good fight and like putting out awesome content off all these years, sir. Appreciate it.