The Ledger
Every statement that passed quotation and attribution checks. Mix any filter with any other: certainty 1/5, debate potential 5/5, or both at once.
why aren't all 5,121 resolved? a statement only gets an assessment when the public record can support or contradict it. opinions and what-ifs never can, and 43 checkable ones are still open, waiting for their date. predictions held up or didn't; assertions are supported or contradicted. on every card: ▮▮▮▮▮ certainty · ▮▮▮▮▮ debate potential. speakers are clickable
Shipper: GPT-5.6 with Codex is best model for knowledge work
“Codex with five, six, which I find to be the most usable, most powerful, fastest model for, specifically for knowledge work and for coding, but like specifically for knowledge work”
Isenberg: In the AI era, users will pay for software maintenance
“Yeah, I think yes, you can build anything, but people will pay for maintenance. That's sort of like the insight is that, you know, is if this thing works, I'm fine to pay for it.”
Dan Shipper: Fable AI autonomously emptied his team's bug backlog
“Like with fable, I was we were like literally The, our bug backlog was emptied out. You could just set it off and it was emptied out as long as you're willing to spend the money.”
Shipper: Anthropic's Claude and Opus remain ahead in visual product design
“I would go to Claude for the actual design, and then port it back in here, because I still think that Claude has a better sense, especially Fable, RIP. But even, even four eight would be better at this. And I think for a product like this, being able to see it…”
Shipper: OpenAI and Anthropic do not really offer fine-tuning anymore
“In that case, like, you have to use open-source models, because none of the major model companies, or at least OpenAnthropic, don't really do fine-tuning anymore.”
Shipper: Frontier AI models cannot copy edit like human copy editors
“These models can't do copy edits. They can't do copy edits like a copy editor can. And it's weird because copy editing shouldn't be that hard. At least copy editing to like 70% or something like that.”
Isenberg: Building fully autonomous AI employees creates bad businesses
“Most people hear agent and imagine a fully autonomous employee. And that's how you get these demos that you see on Twitter and they don't really work and it ends up being a bad business.”
Isenberg: The competitive tech moat is shifting from software to hardware
“Software was an incredible business for the last 20 years. Building SaaS, and there's still opportunities in, in building SaaS, building consumer mobile apps, enterprise apps, but the mode is, is moving to hardware”
Isenberg: The first one-person billion-dollar founder will be a builder-distributor
“I think that, you know, when people talk about the one person, one billion dollar startup, you know, Sam Altman talks about it You know, I think that person is going to be the builder distributor.”
Isenberg: Event opportunities have shifted from massive conferences to bespoke gatherings
“People don't want massive events anymore. They actually want these kind of smaller, more bespoke events, and that's where the opportunity lies.”
Isenberg: Memory Hardware and AI Token Prices Are Not Getting Cheaper
“Because, you know, the price of memory isn't getting cheaper. And the price of tokens aren't getting cheaper.”
Local AI on gaming GPUs handles 80% of typical ChatGPT workloads
“Today a model that runs on a gaming GPU or a decent Mac is good enough for about, I would say, 80% of what most people use things like ChatGPT or Cloud for.”
Hermes is the most used AI agent in the world
“Hermes is the most used agent in the world right now.”
Small local AI models with tools beat giant models without them
“So a small local model with web search, file access, the ability to run code beats a giant model with none.”
Regulated sectors are legally barred from sending data to cloud AI APIs
“Healthcare, legal, finance, they have money, they have problems AI can solve, but they legally cannot send their data to a cloud API.”
Local AI infrastructure acts as insurance against unexpected cloud platform bans
“The lesson is don't build your entire life On something that can disappear with a single letter. Own a part of your stack. Have the generator in the garage. Local models are the insurance, and this is the weekend I finally bought the policy.”
Isenberg: Claude Fable 5 is the most powerful AI model ever seen
“I mean, this is the most powerful model that we've ever seen, and people are just, they're just using it the wrong way.”
Isenberg: AI video editing saves hundreds of thousands of dollars
“This is something that would have cost hundreds of thousands of dollars to make, and using Fable with, ah, to video edit is totally possible.”
Isenberg: Claude Fable at low effort outperforms Opus at high effort
“Fable low is better than opus high, so it's pretty impressive.”
Isenberg: One-prompting startup ideas rarely achieves product-market fit
“So what I'm seeing on X and other places is people are just one prompting these startup ideas. And they're not getting really what they want. It might look visually great, but from my perspective, I don't think it's something that has a high likelihood of PMF,…”
Isenberg: AI focus groups work because sub-agents can maintain distinct personas
“So this is only possible now because sub-agents can hold distinct
Personas and argue instead of a model agreeing with itself.”
Isenberg: Sell local businesses $5,000 custom internal software built in 48 hours
“Every med, spa, contractor, and clinic runs on spreadsheets, group text, because custom software costs like 80,000 dollars a year. So you sell them internal tools for 5000 dollars flat, and you build it in two days. The interview prompt is your sales call. You…”
Mike: AI agent loops are a mistake without money to burn
“You're gonna understand why people are fanning out about it, and you're gonna understand why it is a terrible mistake, and unless you have money to burn, that you are not to do it.”
Mike: Autonomous AI agents usually guess wrong and burn money
“And believe me, when you give the agent the floor to give assumptions, most of the time it's going to get it wrong, but not only is it going to get it wrong, it's going to burn a lot of money.”
Mike: AI can replicate sauce but cannot create sauce
“When you and I are trying to use AI to build something meaningful, I hundred percent stand in the fact that the human still needs to be in the loop. AI can replicate sauce. It can't create sauce.”
Mike: Peter burned $1.3M in AI tokens in one month on loops
“The one argument I'll fight back with is this is going to burn a lot of tokens, and if you don't believe me, all you have to do is look at Peter's tweet, where in one month he burnt 1.3 million dollars worth of tokens.”
Mike: AI agent loops only work in constrained feedback environments
“The only place a loop makes sense is in a very confined, constrained process with a very fixed feedback loop, a very defined feedback loop, and that's in code review.”
Taba: Platforms like Glean and Notion AI create dangerous vendor lock-in
“There's some other companies solving this major enterprise level, other levels like a Glean, for example, Notion AI. They're like search plus context plus an agent layer, chat layer on top. If you want to get locked into that provider, a little bit of a black …”
Taba: AI workflows reduce functional software prototyping from weeks to minutes
“A proposal might've taken up to three days and now it takes minutes. You can saw that clickable prototype again, not a prototype in Figma, not a design prototype. If functional prototype could take one to two weeks to figure out what to build, do it, get it ou…”
Alex Finn: Hermes Overtook OpenClaw with Desktop Release
“I believe this is like the moment that Hermes overtook OpenClaw, to be quite honest”
Finn: Claude Opus 4.8 is the smartest AI model available
“I think Opus for eight is the smartest AI model out there.”
Finn: ChatGPT-5.5 is better than Claude Opus 4.8 for coding
“Coding, I actually think ChadGPT-Five-Five is better than Opus-Four-Eight for coding.”
Finn: Multi-agent role hierarchies create severe token and cost inefficiencies
“The challenge is, I think there's inefficiencies there when it comes to cost. Like, if I have to go to my CEO and say, hey, build me you know, a thumbnail for my latest YouTube video, and then they go to, like, the product manager, who then goes to the graphic…”
Finn: Hermes produces polished updates while OpenClaw updates are rough and untested
“The Hermes updates have been very focused, polished, tested, and tied to really nice narratives that are focused on making life easier for the user. While I feel like the OpenClaw updates Have been more kind of shotgun, put everything in it, don't test it quit…”
Finn: OpenClaw hardcodes models and requires framework updates for new releases
“OpenClaw, like every model is like hard coded in to OpenClaw. So you have to like ask them to switch models. And when new models come out, you have to wait for OpenClaw to update with that newest model.”
Courtney makes €500k from a 1,000-viewer weekly live stream without ads
“I mean, I can just tell you too, like, so I have only one to two, one to 2000 people tuning into me every Tuesday, and that business alone can make four to 500,000 euro with no ads, with nothing else other than me selling like in-person events or like even mer…”
Isenberg: Niche agent-first action apps will be acquired like Instagram was
“Like I think someone can go and create a studio that just focuses on action apps. And I think those, you know, assuming they're picking the right niche, assuming it's a real pain point, assuming the customer has willingness to spend on that pain point. There's…”
Isenberg: Thousands of AI employee agencies will successfully launch
“I think that there's a, there's just, you know, thousands and thousands of companies that are going to come and do this, set this up and there's opportunity for all of them.”
Courtney: Creative workers hate AI for creativity but welcome menial automation
“Everyone in my world hates AI for creative work. It's like everyone's against it, but
Is very positive on AI for sort of menial, annoying tasks.”
Isenberg: AI slop will fail in 2026 while high-quality AI media thrives
“I don't, I'm not a believer that AI slop is going to work in 20, 26 and beyond. Like, I think that that's frankly a waste of time, but I do think that using AI to create media companies that are actually high quality, maybe with some, with a human in the loop …”
To: Traditional UI templates are incompatible with AI coding agents
“You can go to any of the templates that you find on VZero, Lovable you know, all of these other place Framer, Figma community for people who are more familiar with designs, and the problem with those is that they are highly close behind the techniques and the …”
To: AI builders must rapidly adopt new models to maintain moats
“As someone who works in AI, your mode is basically, you have to be new, you know, you have, like, I don't, I'm sure anyone can relate, But you always, when there's a new model that comes, you have to talk about it. Otherwise you're kind of losing a lot of that…”
To: Products built purely with AI will show visible defects
“They want real people, they want real craftsmanship behind all of these pixels at some point, because, you know, the cracks are going to show if you only use AI. The, you know, the whole way.”
To: Derivative product design slashes perceived value by up to 100x
“Niche is the key word here because you cannot compete anymore with people who are generalized. It's too difficult. Nowadays we want authenticity, we want niche, we want taste, and if something looks like another thing, it kind of reduced the value by, like, al…”
Wilkinson: Creativity is compromised by the number of intermediaries in execution
“And I think that creativity is just compromised based on how the number of people between your vision and execution.”
Wilkinson: 5M to 10M context windows will enable AI-run companies
“And I feel like once the context window is like 5,000,010 million, then we'll be able to run entire companies with this.”
Isenberg: Fully autonomous companies are still an unrealistic dream
“You cannot run a fully autonomous company right now that, yeah, that is going to get better, that you can just like, You know, go hang out on the beach, 24 seven. That being said, I think there's parts of the business that can be autonomous. Like support is li…”
Wilkinson: OpenClaw agents still require micromanagement, not true delegation
“Like I think open claw agents are still like zap. They're basically like zap your zaps, but that can make basic decisions and have intelligence, but you still have to tell them step by step. This is how I want you to think. This is how I want you to operate. L…”