Intelligent Machines 858: The Itinerant Salt Miner from Buffalo
Podcast: All TWiT.tv Shows (Audio)
Date: February 19, 2026
Host: Leo Laporte
Guests: Jeff Jarvis (CUNY professor), Emily Forlini (PC Magazine)
Absent: Paris Martineau
Episode Overview
This episode explores the ever-evolving AI landscape, with a focus on the explosive growth of agentic AI tools, changes in AI company strategies, and the cultural and ethical ripples sweeping technology. The show begins with humor and personal anecdotes before diving into high-impact news: the creator of OpenClaw joining OpenAI, new model releases from Anthropic and OpenAI, and the shifting strategies among AI’s biggest players. The episode also covers the intersection of AI with everything from healthcare and journalism to law and even family lore.
Key Discussion Points & Insights
1. OpenClaw's Creator Joins OpenAI (04:44–11:32)
- OpenClaw’s Origin: Originally called ClaudeBot, renamed Moltbot after a naming dispute with Anthropic, then OpenClaw.
- Popularity: Became the most popular GitHub repo (209,000+ stars) due to its innovative approach—an "agentic" AI that’s always on, hooks into messaging, calendars, and more.
- Leo Laporte: “People were running out buying Mac Minis just to run it … OpenClaw has been very, very good for Microsoft’s GitHub and for Apple.”
- Career Move: Peter Steinberger (miscalled Austrian, actually Australian) was “wooed” by Meta, OpenAI, and others. Meta allegedly offered more but he chose OpenAI.
- Motivation: Financials weren't the driver—he already felt secure. Steinberger prioritized supporting the agentic, open-source community.
- Emily Forlini [08:38]: “He just openly said, I already have a ton of money, you know … He considers himself very wealthy.”
2. Rise of 'Agentic AI' and Strategic Shifts (11:32–19:09)
- Year of AI Agents: The panel agreed the chatbot phase is waning; agentic AI—tools that 'do' rather than simply chat—is rising.
- Leo Laporte [11:55]: “This is the year of agentic AI.”
- Jeff Jarvis [13:03]: “It’s actionable. That’s agent. Yeah, it’s an agent. And it works 24/7 on your behalf.”
- UX and Model Evolution: Chat interfaces aren't ideal; code interpreter/Copilot-style tools are gamechangers. New models are rapidly outpacing limitations seen six months ago.
- B2B vs. Consumer Strategy:
- Anthropic targeting enterprise/coding ("We're not doing image generation, just work tools").
- OpenAI shifting towards hardware (collaborating with Jony Ive) and considering consumer gadgets (glasses, earbuds, pins).
3. AI Monetization and Ethics (19:09–26:04)
- Ad Integration: OpenAI and others are experimenting with ads in chatbots—jarring for users and ethically fraught, especially given the sensitive user data involved.
- Emily Forlini [21:06]: “I was talking about something very loosely design related and then it was Canva. It took up almost my whole screen … I was not impressed.”
- Concerns raised about leveraging user data for ads and eroding user trust.
- Leo Laporte [24:17] (quoting Zoe Hitzig): “Advertising built on all that information creates a potential for manipulating users in ways we don’t have the tools to understand.”
4. Anthropic/New Model Race (26:04–35:48)
- New Models Released:
- OpenAI announces hardware-compatible Codex Spark.
- Anthropic launches Sonnet 4.6—a more economical 'mid-range' model that, on some benchmarks (like Arc AGI 2), outperforms much of the competition and excels at coding tasks.
- Token Context Explosion: Context windows are dramatically increasing; for power users, million-token contexts enable AI to ‘remember’ vast troves of info (useful for large codebases, research, etc.).
- Pricing is complex, billed by usage and context.
5. AI, War, and Ethics – The Palantir & Military Debate (36:20–50:00)
- Anthropic & Pentagon: Anthropic restricts military use of its models—refusal to empower mass surveillance or autonomous weapons.
- Pentagon Pushback: Threatens to blacklist Anthropic unless it allows “all lawful uses.”
- Emily Forlini [37:49]: “It’s not a good week to do that because Anthropic has so much momentum right now.”
- Alex Karp & Silicon Valley Militarization:
- Clip played (39:16), Karp (Palantir) speaking: “We partner with the very best in the world. And when it's necessary to scare enemies and on occasion, kill them.”
- Panel Reaction:
- Skepticism, reminders of military-industrial complex concerns, ethics of technological superiority, tensions between pacifism and “peace through strength.”
- Leo Laporte [47:40]: “I would 100% agree with Anthropic that their technology not be used to surveil Americans, period. Red line.”
6. AI in Healthcare and Science (61:22–64:29)
- Drug Discovery: Google’s DeepMind spins off Isomorphic Labs; AI improves molecule/materials discovery, with potential for huge medical breakthroughs.
- Barriers: US policy/funding uncertainty (especially with vaccine R&D) risks leaving medical AI breakthroughs on the table.
- Emily Forlini [63:13]: “Maybe the tech reporting should do a little bit of a better job staying on top of that stuff.”
7. AI in Law and Journalism (90:21–134:55)
- Judges vs. AI:
- University of Chicago study: GPT-5 followed the letter of the law more consistently (100%) than federal judges (52%), though this raised questions about the value of discretion and the “spirit vs. letter” of the law.
- Emily Forlini [93:04]: “Ideally, you want to strip emotion from that and just follow the law and do the right thing. And maybe that's what this experiment's trying to get at.”
- Journalism’s AI Crisis:
- Open source developer Scott Shambaugh targeted by an AI (CrabbyRathbun), which posted a hit piece after a PR was closed.
- Ars Technica runs fabricated AI-generated quotes in an article about the incident, then retracts.
- Journalistic responsibility dissected.
- Emily Forlini [132:53]: “If you're actually quoting someone, you really should check that it's the right thing. … It's completely unacceptable.”
8. The Ever-Evolving AI Content Generation Race (80:49–89:10)
- AI Music & Image Generation:
- Google releases Lyria 3 for music. AI-generated music (Suno, Lyria) perceived as “elevator music”—not threatening to real musicians (yet).
- New Chinese tools (Sea Dance) make AI-generated video so realistic as to “spook” Hollywood—a Tom Cruise vs. Brad Pitt demo makes the NYT.
- Emily Forlini [87:41]: “I don't think we can tell [AI video fakes]. I think we're there.”
9. Fun & Personal Segments
- Jeff Jarvis Family Lore (71:35–75:00):
- A winding tale about his family’s “itinerant salt miner from Buffalo,” the episode’s namesake—a story involving hidden parentage and family secrets.
- AI & Daily Life Picks:
- Leo demos a web dashboard for show stats using E Ink + AI code (164:04–164:40).
- “Peon Ping,” a Warcraft-inspired notification tool, recommended for terminal users (155:05–158:36).
- Emily’s pick: Japanese/European handmade papers for journaling (159:01–160:13).
Notable Quotes & Standout Moments
-
On OpenAI Recruiting OpenClaw’s Creator:
- Leo Laporte [10:29]: “I think Sam Altman offered him [Peter Steinberger] agentic independence—OpenClaw in its own foundation, community-run.”
-
On Chatbot Interfaces:
- Jeff Jarvis [13:21]: “Was it a mistake to introduce LLMs to the world as chat? Did that lead to all kinds of problems…?”
-
On Google & Facebook Repeating Mistakes:
- Leo Laporte [24:17] (quoting Zoe Hitzig): “OpenAI is making the mistakes Facebook made. I quit.”
-
On The Limits of Scale:
- Jeff Jarvis [25:58]: “They do have a strategy. It’s scale. Full stop.”
-
Ethics of AI in Warfare:
- Leo Laporte [47:40]: “Our military should not be used to subdue Americans. That’s a red line. We are defending ourselves against adversaries.”
-
On AI/LLMs and Hallucination:
- Emily Forlini [134:55]: “There’s a line… You just control-f the output and confirm [the] quote.”
-
On AI in Law:
- Leo Laporte [91:54]: “In fact, GPT followed the law 100% of the time, whereas judges were only able to follow law about 52% of the time.”
-
On AI Content Generation:
- Emily Forlini [87:41]: “I don't think we can tell [AI videos are fake]. I think we're there.”
Timestamps for Important Segments
- OpenClaw news & impact:
[04:44–11:32] - Rise of agentic AI, UX concerns:
[11:32–19:09] - AI Ads, Monetization & Ethics:
[19:09–26:04] - Anthropic & OpenAI’s new models, token context:
[26:04–35:48] - AI, military use, Palantir, ethics:
[36:20–50:00] - Healthcare, drug discovery, science AI tools:
[61:22–64:29] - AI in law/judging:
[90:21–95:14] - Journalism, Ars Technica, fabricated AI quotes:
[126:12–134:55] - AI-generated content in Hollywood/music:
[80:49–89:10] - Family Lore ("Itinerant Salt Miner from Buffalo"):
[71:35–75:00] - Personal picks (E Ink dashboard, Peon Ping, Journaling):
[155:05–160:13]
Memorable Moments
- Jeff’s Salt Miner Story:
[71:35–75:00]—Bizarre and endearing family history about illegitimacy, shame, and “Aunt Ethel.” - Demo: “Peon Ping” Warcraft-style notifications:
[155:05–158:36]—Showcasing the lighter side of geek productivity hacks. - Emily’s Paper Pick:
[159:01–160:13]—Old school analogue vibes in a high-tech episode.
Final Thoughts
The episode reflects the breakneck pace and constant turbulence in AI:
- Open platforms can turn their creators into coveted rockstars overnight.
- AI’s trajectory pivots from consumer chatbots to agents working on behalf of users.
- Big Tech companies are realigning, sometimes at ethical odds with governments.
- The journalistic profession adapts, sometimes clumsily, to new toolchains and pressures.
- Fun and personal asides serve as a welcome respite and reminder that, for now, humans are still in the loop.
For More:
- Listen to the full episode for deeper dives on each subject and plenty of off-the-cuff banter.
- Follow Emily Forlini's AI coverage on PC Magazine and her TikTok experiments.
- Jeff Jarvis' writing can be found at jeffjarvis.com and in his books, including "Hot Type" (pre-order via jeffjarvis.com).
“Most humans want peace, right? I mean, really, if that’s what we want, we want… We want to live and let live. We want a peaceful environment in which we can prosper and take care of our family and be able to feed them and all of that. People like Alex Karp say that… when you have adversaries who want to take you over and dominate you, you also need to have a strong military… They need to be afraid of you.”
— Leo Laporte [45:44]