Welcome back! LinkedIn’s “Seems like AI slop” button has already been clicked more than 1 million times, which feels less like a product update and more like a cry for help.
The platform says users are now seeing 40% fewer views on posts it classifies as AI slop, and it’s adding a warning for people whose posts get flagged. Somewhere, a LinkedIn thought leader is realizing the em dash was a liability.
Would You Let ChatGPT Watch Everything You Do?
OpenAI just rolled out a new feature called Computer History, and it reminds me a lot of Microsoft Recall. Which, if you remember, everyone hated.
Here's the idea. ChatGPT and Codex can now watch what you're doing across apps and websites on your Mac and turn all that activity into a searchable timeline. Then you can ask things like: "What was I debugging yesterday?" or "Review everything I did today—what parts of my workflow could you automate?"
The differences from Microsoft Recall are the whole story:
No screenshots. Recall took constant screen captures. Computer History uses macOS accessibility APIs to record structured events—clicks, typing, app switches. Much less blunt (and notably, it replaces an earlier OpenAI experiment called Chronicle that used screenshots, which OpenAI clearly learned from).
Opt-in by default. Has to be manually enabled. Microsoft famously got roasted for shipping Recall on by default.
Granular controls. Choose which apps get tracked, pause from the menu bar, delete individual entries. Incognito browsing is auto-excluded.
Data policy. OpenAI says it processes events into text memories, then discards the raw event files, and won't use them for training.
Currently Mac-only, requires a paid plan (Pro, Business, or Enterprise), and not yet available in the EU, UK, or Switzerland.
My POV: On one hand, giving an AI access to essentially everything you do on your computer sounds like a privacy nightmare, and there's no getting around that framing. OpenAI clearly learned real lessons from the Recall backlash. But we're still talking about your entire digital workday being logged into a database at an AI company.
On the other hand, this might be one of the most powerful ways to actually figure out what AI can automate for you. Pair Computer History with something like Codex's Record & Replay feature (which learns a workflow from watching you do it once), and the natural next step is Codex spotting a repetitive workflow on its own and offering to build the automation. That's the agentic loop starting to close.
For power users doing serious knowledge work, I think a lot of people will end up turning this on. For everyone else, the privacy tradeoff will feel like too much for too long.
Hit reply and let me know, would you turn this on?
— Matt
Hugging Face Reportedly Fields $13B Acquisition Interest
Via Business Insider
Hugging Face has reportedly been approached about a sale that could value the AI developer platform at $13 billion or more. No deal has been reached, and we don’t yet know who the potential buyers are.
Platform layer: Hugging Face has become one of the central places where developers and researchers share, discover, test, and deploy AI models. Rather than competing directly with OpenAI or Anthropic, it sits underneath much of the AI ecosystem as a distribution and tooling layer for builders.
Jump in valuation: The reported $13 billion figure would be a major jump from Hugging Face’s last known valuation of $4.5 billion in 2023. Its investors include Salesforce Ventures, Alphabet, GV, IBM Ventures, Lux Capital, and Addition.
Missing details: The talks may not mean Hugging Face is actually eager to sell. CEO Clément Delangue has emphasized the company’s long-term responsibility to its community. Hugging Face reportedly turned down a $500 million Nvidia investment earlier this year that would have valued it at $7 billion because it did not want one dominant investor influencing decisions.
The bigger picture: Stripe’s reported OpenRouter deal showed how valuable model-routing infrastructure has become. Hugging Face would be an even bigger prize: a community, a distribution layer, and a critical piece of AI’s open model infrastructure all in one. A buyer might be eager, but the question is whether Hugging Face is willing to sell.
OpenAI Wants California to Strengthen Its AI Safety Bill
OpenAI is calling on California to add stronger safeguards to SB 53, the state’s landmark AI safety bill that passed last year.
Quick refresher: SB 53 created transparency requirements for large AI companies along with whistleblower protections. The bill was designed to give California more visibility into frontier AI development and create basic accountability measures as powerful models become more capable.
What OpenAI wants: OpenAI wants expanded safeguards, including requiring monitoring of frontier models during training or evaluation for potential serious incidents and strengthening cybersecurity protections throughout the model-development lifecycle. OpenAI pointed to recent incidents as evidence that AI safety rules need to keep evolving.
The reversal: The endorsement is notable because OpenAI previously opposed SB 53. The company now says that, without major federal AI legislation, states can move in a compatible direction around core protections that could eventually form the basis of a national standard.
The bigger picture: OpenAI’s change of tone shows how quickly the politics of AI safety are changing: companies want rules that reassure the public and enterprise customers, but they also want a seat at the table while those rules are written.
Analyze Landing Pages Before Buying Traffic
Via LandingBoost
LandingBoost is a landing page analysis tool for SaaS founders that scores pages across four conversion axes: clarity, relevance, trust, and action. It identifies the single highest-impact edit to make before sending more traffic to the page, then gives users a paste-ready implementation prompt.
How you can use it
Score landing pages for clarity, relevance, trust, and action
Find the highest-impact edit to make first
Get a paste-ready prompt for your site builder
Monitor pages for conversion regressions
Pricing: Paid
Project Management for Humans and AI Agents
Via TaskFolk
TaskFolk is a project management platform that combines boards, backlogs, sprints, timelines, docs, goals, reports, and automations in one workspace.
How you can use it
Manage boards, backlogs, sprints, timelines, docs, and goals
Delegate tasks to human teammates or AI agents
Track agent work inside the same project workspace
Connect AI agents through REST API and MCP integrations
Pricing: Free and paid
Jobs, announcements, and big ideas
Anthropic unifies Claude's memory across Chat and Cowork, giving users control over which topics it keeps.
OpenAI ships an admin plugin for ChatGPT Work and Codex.
Bill Gates warns “bad stuff is imminent” with AI.
Stability AI lands a $76M Series B backed by EA, Sony, Universal, and Warner Music.
Perplexity debuts a local-first AI agent that keeps private knowledge work entirely on-device.
SpaceX plans to build a $100 billion launch complex for its Starship rocket.
Google teams up with Delaware to offer free AI training and career certificates.
Apple unveils its 2nm M6 and M5 Ultra chips, promising major leaps in AI performance.
What if AI could build worlds? Tencent’s WorldClaw turns a prompt into a fully editable 3D environment, object by object ⤵️
That’s a wrap! See you Friday for more.
