Sam Barnett

"And why do we fall? So that we can learn to pick ourselves up.”

@chrisman “How can I tell what I think till I see what I say?” — E.M. Forster

So many comments just asserting “but writing is good!” Try a Socratic dialogue with your 9yo. You will be able to help them discover and correct flaws in their thinking about 100x faster than if you first had them go through the laborious effort to transcribe them by hand.

Making kids write their ideas is just an unfortunate side effect of large classrooms. If you homeschool, you can just talk with your kids. Far more effective. Neither Socrates nor Jesus felt compelled to write persuasive essays. Probably 9 year olds don’t need to either.

@chrisfralic Framework + example

I’m looking for examples of well written cold emails that really landed - they got a response and maybe it led to a meeting or maybe something much bigger. Please copy or point to them in the replies - thanks! 🙏

@KobeissiLetter @grok what companies are building out all this pipeline?

US natural gas pipeline construction is skyrocketing along the Gulf Coast: The Gulf Coast is set to add ~18 billion cubic feet per day of new natural gas pipeline capacity in 2026, the biggest one-year increase in 18 years. This represents a +13% jump in the region’s total capacity, the equivalent of Canada’s entire consumption, driven by 12 pipeline projects across Texas, Louisiana, and Oklahoma. In the past, the only larger expansion occurred at the peak of the shale-gas boom in 2008. The US' energy dominance is entering a new era.

@rexsalisbury @TurnerNovak If that were true what’s the point of tokens?

@SamRBarnett @TurnerNovak Consumers / Users is a reasonable starting point. Deep agentic usage is additional upside.

AI “bubble” needs $400 / user / year to justify current capex by 2030. For comparison in US - Apples makes $1k / year for 165 million users. - Google also around $1k / year for ~250 million users . - Facebook around $240 for ~250 million users. $1k / year is also rough cost most spend on cell plans per year. $400 is a lot but not unprecedented.

@TurnerNovak @rexsalisbury ‘Per user’ is not the right way to scope consumption. Users are not a relevant proxy for agents. Multiple agents can act on behalf of a user. Agents that reason and use tools will represent the majority of token consumption. The framing of this analysis is not very relevant.

@rexsalisbury This is all consumer, no B2B?

@tbpn @EverettRandle Definition of ‘agent’ is clear. An agent is the combination of: (1) instructions / prompt (2) tools used to support instructions / prompt (3) data used to support instructions / prompt

.@EverettRandle works through when we'll get AGI: "AGI has become a near-useless term. It's almost like 'agents' at this point. Both are so nebulous, and the product marketing around them has been so brutal, that they've lost all their meaning."... As for when we'll get AI that can do what a reasonable adult could do in any given situation, "it's fully dependent on how quickly we can distill the technology through the economy. It takes a really long time, no matter how fast [AI company] growth curves are."

@arpitrage @arpitrage - open source modes do not offer indemnification. Open source users must indemnify the provider. Consumers may be comfortable but enterprises are not

Bearish for the AI bubble: - Chinese tech firms can train frontier models pretty cheaply - They shoot straight to the top of the leaderboards, hugging face downloads, are open source - Get incorporated in wrappers like Perplexity Shows US foundational models have no moat

@tgeisenheimer @jasonlk @bhalligan If an agent requires many tools to complete a task/set of tasks (ie deliver an outcome), then the token consumption can be significant. The outcome may not be super valuable. The cost of the former (consumption) does not necessarily equate to the value of the latter (outcome).

@jasonlk @bhalligan Why is outcome based pricing meaningfully different than usage based pricing?

I was really excited about the outcome based pricing model à la what Sierra, Fin, etc are doing, but the more I look into it, the more I think it won’t become the norm. Too hard to apply in 99% of businesses. Am I wrong?

@mwseibel Do you think this is also true for revenue you help your customers retain? Arguably it is more expensive to replace an existing customer than to keep one and grow.

Often founders think they have a bad GTM motion or need a new sales team, when the reality is their product doesn’t help their customers make enough money.

The amount of money your company can make is directly related to the amount of incremental revenue your customers believe you help them make. Revenue your customers believe you help them make is therefore a better KPI than your revenue because you can always figure out how to optimize your pricing over time. And your customers will almost never churn as long your product continues to help them make a lot of money. Let me know if you can find an example at scale (companies with over $1b in revenue) where this is not true.

@shai_wininger @shai_wininger have you considered using @withgraphite for code review?

Validation remains the main bottleneck in agentic coding. As agents generate features quickly, humans are left to review and validate their work: security, stability, performance, compliance, UX, design, copy, and more. This is not scalable. Truly unattended, production-ready agentic coding requires rethinking how quality is built into the process. Our approach is to embed validation at the very beginning of feature creation. We've built AI tools that analyze product specs, generate test plans, orchestrate complex testing with natural language, and emulate human interaction to "try" new features. The @Lemonade_Inc AI Validation Framework is still in its early days, but we're already seeing strong signals of scalability and impact. In July, our tiny QA team ran 82,167 complete quality tests, covering 95% of customer-facing production code. Using traditional methods, this would likely have taken 10x the headcount and 10x the time. Below is a small part of the system in action:

@mvcinvesting twitter.com/jrichlive/stat…

Is the current pace of AI data center construction a bubble, or is it the building of the necessary infrastructure to support the most transformative evolution of the 21st century?

Mentorship Moment #52

@tbpn @mcuban Innovative companies are hiring for this role in house x.com/adamdangelo/st…

Mark Cuban (@mcuban) on the next big job students should focus on. Most companies don’t know how to implement AI, especially small businesses. “There is nothing intuitive for a company to integrate AI.” “Companies don’t understand how to implement AI right now to get a competitive advantage… learn to customize a model, walk into a company, show the benefits. That is every single job that’s going to be available for kids coming out of school.” Don’t just study AI. Make it work inside a business.

@venturetwins @a16z @arampell Or one app to rule them all? semianalysis.com/2025/08/13/gpt…

🚨 New @a16z thesis: AI x commerce AI will change the way we shop - from where we find products to how we evaluate them, when we buy, and much more. What types of purchases will be disrupted, and where does opportunity exist in the age of AI? More from me + @arampell 👇

🚨 TRUMP JUST SAVED COLLEGE SPORTS. No more billionaire bidding wars. No more boosters buying rosters. No more chaos. Trump just signed an executive order banning the “pay-for-play” deals that turned college football & basketball into free agency. Here’s what it means 🧵

@levie twitter.com/omarsar0/statu…

Context engineering is increasingly the most critical component for building effective AI Agents in the enterprise right now. This will ultimately be the long pole in the tent for AI Agents adoption in most organizations. We need AI Agents that can deeply understand the context of the business process that they’re tied to. This means accessing the most important data for that workflow, using the appropriate tools at the right moment, having proper objectives and instructions, and understanding the domain that they’re in. Some of the big open items for anyone building enterprise agents are: * Narrow vs. General agents. The smaller the task, the easier it is to give the AI Agents the right context to be successful. But the smaller the task, the less value there will be. Finding the optimal task size for value generation will be an important factor for the next few years. * Getting data into an agent-ready system. Enterprise data is often fragmented between dozens or hundreds of systems, many of which are not prepared for a world of AI. Most companies will still need to modernize their data environments to get the full benefit of AI Agents. * Accessing the *right* data for the task is paramount. Even when you have data in a modern environment, getting access controls perfectly aligned to what the AI Agent is going to need access to is critical. Further, deciding what to do RAG on vs. just a general search vs. what to put fully into the context window will matter a ton per task. * Choosing what should be deterministic vs. non-deterministic. If you demand too much from the models, you’re likely to see some drop off in quality. Yet, if you have the model do too little, then you’re dramatically underutilizing what’s possible with AI. This of course is a moving target because the models themselves are improving at an accelerating rate. * The right user interface to get the AI Agents context deeply matters. Half of the problem for getting context to agents doesn’t look like an AI problem at all. It’s all about where the agents show up in the workflow and how the user interacts with them to provide them the context necessary to do the task. The race for the next few years in AI in the enterprise is to see who best to deliver the right context for any given workflow. This will determine the winners and losers in the AI race.

@lennysan JD posted ~1 month ago at Quora twitter.com/adamdangelo/st…

Why you should probably hire an "AI Operations Lead," whose job is to make everyone at your company more productive using AI tools and workflows. "I sit with her once a week, and every time I'm doing something repetitively, we put it in a to-do list. She then builds prompts and workflows so that I and everyone else on the team are automating as much as possible. If you're working in a job all day, you're fighting fires, you're like, okay, am I gonna do this in the way that I know how, or am I gonna do it in the new way that might not work? Having an AI ops lead lets, you basically identify those things and then have them solved, without people who are doing the work having to take time to do it. Which makes it much more likely it happens." — @danshipper

@ecommerceshares Check out Cohere: cohere.com/private-deploy…

There’s a huge opportunity for a privacy-focused LLM provider. Because it doesn’t exist today. It’s a wide open field. Everyone wants to be the next Zuck & sell your data to the highest bidder. But people in medicine, business, finance, defence etc don’t want their data sold.

@cixliv Ran right past an excellent burrito spot

I guess a video of me running the robot through SF yesterday is going viral on TikTok

one of Steve Jobs' best lines. simplicity is so hard.