TLDR Founders 2026-09-07

micro1 partners with companies to license operational data for training and evaluating frontier AI models.If your company has 30+ employees and years of data showing how work actually gets done, including SOPs, project histories, knowledge bases, CRM data, QA processes, or internal workflows, it could qualify for an Enterprise Data Partnership.Businesses can earn$100K-$2M+ depending on the volume, quality, and uniqueness of their data.Your company retains ownership of its underlying data, and all data is securely transferred, de-identified, masked, and transformed before onward licensing for AI training.

AI will make systems of record more important as customers will soon be able to pipe the system's data into an AI agent and skip the AI-native app. An incumbent system of record working with a general agent will be enough to get the work done. The incumbents have an incentive to control and charge for access to their data, and also to push their own agents forward. The vertical AI-native company has to perform a specific cross-system job better than an incumbent's purpose-built agent or a general-purpose agent that can reconstruct it from the underlying tools.

California signed SB 122 on June 29, which will apply sales tax to SaaS and AI tools starting on January 1. The rate is 7.25% at the state level plus local district taxes, so the total lands between 7.25% and 10.75% depending on where the buyer sits. Sacramento projects about $2B a year. Remote sellers cross into collection duty at $500,000 in California sales. Every California customer's invoice grows by up to 10.75% that day, and the vendor collects it, so pricing pages, billing systems, and renewal conversations all change before year-end.

A small e-commerce analytics company compared over 2 million product listings across more than 100,000 search result pages and AI Mode answers over 23 days in August, covering the US and UK. Products that appeared in both surfaces were 21.6% more expensive in AI Mode, with a median price of $149 against $100 in traditional search. Only 1.28% of products in traditional results also showed up in AI Mode for the same query, so a rank in traditional search says almost nothing about whether a product is in the answer box. The main seller differed on 49.6% of matched products.

Announcing the behavior you want does approximately nothing. You get the behavior you reward. This is true anywhere people watch each other for signals about what counts. It matters most when trying to fix something in the culture.

A one-person company can run recurring work through Slack by giving one AI assistant shared business context, connected tools, scheduled briefs, and strict approval gates. The setup works best when workflows are added gradually and corrections become standing rules.

Your first customers are most likely already in your network. Start with ten warm prospects, make a direct offer with a price and deadline, deliver manually, and use ten real sales conversations to validate or reject the idea quickly.

Project HydraFusion delivers frontier intelligence through runtime orchestration. It can create a full execution plan, choose from models across multiple providers to draft, critique, and revise, or cascade to more powerful models to complete tasks. Developers just select HydraFusion like any other model, and it chooses a workflow that balances performance, cost, and latency for each task. HydraFusion is now available to users on all GitHub Copilot plans as a research preview.

Zenith is a ready-to-code, distraction-free experience on Windows on developer-class devices. It will first become available with AMD's Ryzen AI Halo. Zenith devices will come with a preconfigured Windows setup for development and a set of tools curated for what developers need. They will be capable of running 30+ parameter models locally and unmetered.

Industrials made up 23% of the 245 companies launched so far in YC S26. The previous five-year average for industrials was just 6.6%. SaaS has fallen to 12% vs a 28% five-year average. B2B software overall is at its smallest share since S22 at 52%. AI appears to be on a downward trend, but this may just be due to companies no longer tagging themselves as such as it has become a default.

Tokens are getting so cheap in China that businesses give them away as loyalty rewards. Daily token use there reached 500 trillion this year, up from 100 billion in early 2024. Chinese models cost 60% to 90% less than American ones. China Telecom sells 10 million tokens a month for about $1.40. A Beijing dumpling restaurant hands out more than 100 compute vouchers a day, and a Shanghai bank bundles up to 3 billion Qwen tokens with a credit card, in the slot where airline miles used to go.

Nearly all startup ideas can be duplicated - their value is the other ideas they lead to.

Charging per seat is slowly making less sense.

Your buyers are asking ChatGPT what your product costs, and it is mostly not reading your pricing page to answer.

Most founders try every acquisition channel at once and never learn which one works.