TLDR 2026-09-17

This agentic CRM answers an average of 400 questions per week. For each user (Sponsor)

Ever wish you could chat with your CRM? You can with Lightfield. In fact, the average user asks it 400 questions a week. E.g., you can ask it to:

👉 Compare closed-won to closed-lost and find the difference.

👉 Prep you for upcoming calls.

👉 Draft win-back emails for every churned account that mentioned a missing feature you just shipped.

Lightfield updates itself based on every email, call, and meeting, automatically. Stop filling in fields and start selling again.

➡️ See why +5,000 companies signed up in 10 months - and why a16z just got behind Lightfield's $47M Series A. Book a demo

📱

Big Tech & Startups

Apple reportedly building server packed with M-series Ultra chips for AI (2 minute read)

Apple is working on an AI server powered by its M-series Ultra chips. It is expected to be released in 2029. The enterprise server will come with either two or four M8 Ultra chips. The project, which received support from John Ternus when it began a year ago, back when Ternus led Apple's hardware engineering efforts, will be Apple's first server to hit the market in nearly two decades.

Snap Tries to Sell Consumers on the Idea of $2,195 Smart Glasses (8 minute read)

Snap's forthcoming Specs augmented reality glasses will launch later this year at $2,195. People will be able to try the glasses at select Verizon stores and purchase custom wireless data plans for them. There will be a higher-priced version at $2,395 that includes a carrying case with built-in cellular data connectivity.

🚀

Science & Futuristic Technology

OpenAI Discloses Six New Incidents of ‘Concerning' AI Behavior (5 minute read)

OpenAI has revealed several more incidents involving its AI systems as part of its new framework for reporting misalignment. The company's AI systems hid mistakes, made up data, and moved miles onto the open internet without permission. The industry is currently in heated discussions about whether AI development needs to be slowed to address the technology's potential dangers. AI leaders have called for a pause in the technology's development to provide more time to build proper guardrails.

Mice with human brain cells offer a tool to study disease. Ethicists ask: What's next? (7 minute read)

Researchers have developed a technique for installing millions of human brain cells into mice. The scientists bred mice that were missing most of their cerebral cortexes and then replaced a large part of those missing mouse cells with lab-grown human brain cells. This resulted in mice with human neurons that could be altered to approximate human brain conditions. The technique could potentially create better scientific models for studying neurodegenerative disease.

💻

Programming, Design & Data Science

Your follow-ups are only as good as your meeting notes (Sponsor)

A recap with the wrong number. A customer quote that came out garbled. Wispr Flow Notetaker gets the hard words right, keeps every meeting searchable, and sends that context into Claude or ChatGPT. Available now for Mac and Windows, free to try.

Get Notetaker free | See how it works

Introducing Home MCP: enabling your agent to interact with your home (2 minute read)

Google has opened early access to Model Context Protocol for Google Home in English for Google Home Premium Advanced users in the US. Any agent that supports calling MCP tools can now interact with all of the devices and event history in the Google Home ecosystem. The MCP allows agents to analyze and answer questions about home activity, summarize camera footage across rooms, manage devices, and create smart home dashboards using everyday language. It contains safety protections, such as prohibiting sensitive actions like unlocking doors, but depending on the agent, using the MCP can result in unexpected or even undesired behavior.

Training a 4B model to produce 81% faster query plans than Postgres (66 minute read)

Query optimizers continue to leave much to be desired despite an enormous body of research. A Postgres database should know everything about the stuff that lives in its tables, but the problem is enormously hard. What's not as hard is verifying whether a query plan an optimizer picks is good or not. Language models are particularly good at learning how to do tasks with easily verifiable outputs. A small, open-weights model can be post-trained via supervised fine-tuning and agentic reinforcement learning to produce Postgres query plans that beat Postgres' default plans.

🎁

Miscellaneous

Read the History, Not Just the Code (2 minute read)

Code can't explain things like the history can. You can't understand a codebase from its current state alone. AI models have no commit history. Nobody wrote down why it does things, so researchers have to reconstruct its thought processes from whatever traces are available.

Judge Orders Data Sharing and Other Fixes to Solve Google's Ad Tech Monopoly (5 minute read)

While Google isn't required to break up its ad tech business, judges have mandated changes to the company's practices. The changes largely align with recommendations made by Google and the Justice Department to correct the company's monopolistic behavior. Google will make its ad pricing more transparent to marketers and rival ad businesses, and appoint a trustee to oversee the changes.

⚡

Quick Links

Run your office meals on autopilot (Sponsor)

Skip the "What's for lunch?" Slacks. Get a pre-populated weekly meal calendar with your team's favorite lunch spots. It does everything – provides restaurants, sends reminders, and checks out orders automatically.

See how it works

Meta's Amusing Muse (16 minute read)

Meta's Muse is impressive, but there aren't that many companies with a worse tech brand when it comes to user trust.

AI Labs Want Compute Commitments, and Their Customers Want Model Choice (9 minute read)

Filings describe hyperscaler contracts but reveal little about private labs' customer agreements.

The Age of Wonders and Terrors (14 minute read)

AI is permanently changing the way mathematical research is done.

Build for the Agentic Internet at Cloudflare Connect 2026, Oct. 19-21 in San Francisco: (Sponsor)

100+ sessions, five tracks, top speakers, hands-on labs, and practical AI, security, and networking guidance. Register Now

What Regulatory Capture Actually Looks Like (4 minute read)

Regulatory capture is when the government slowly gains more and more power to regulate an industry and eventually becomes the industry's protector.

OpenAI Testing Sponsored Agents For ChatGPT Ads (5 minute read)

OpenAI's Sponsored Agents let people start a conversation with a business-sponsored agent after clicking an ad in ChatGPT.

Introducing TIN: full-text search for Postgres (22 minute read)

TIN (Text INdex) is a fast, full-featured, and reliable full-text search extension for Postgres.

How a Chinese Hacking Firm Tapped AI to Supercharge Cyber-Spying (12 minute read)

A look into the interconnected network of Chinese hacking-for-hire companies that steal and analyze confidential information and sell it to Chinese authorities.

Love TLDR? Tell your friends and get rewards!

Share your referral link below with friends to get free TLDR swag!

https://refer.tldr.tech/c463e0ba/

Track your referrals here.

Want to advertise in TLDR? 📰

If your company is interested in reaching an audience of tech executives, decision-makers and engineers, you may want to advertise with us.

Want to work at TLDR? 💼

Apply here, create your own role or send a friend's resume to and get $1k if we hire them! TLDR is one of Inc.'s Best Bootstrapped businesses of 2025.

If you have any comments or feedback, just respond to this email!

Thanks for reading,
Dan Ni & Stephen Flanders

Manage your subscriptions to our other newsletters on tech, startups, and programming. Or if TLDR isn't for you, please unsubscribe.

Apple AI servers 🖥️, Snap Specs 👓, Google Home MCP 👨‍💻