Sign Up · Advertise |
|
|
Welcome, humans. |
Relationship therapist Esther Perel just conducted a couples therapy session between a man and his AI "girlfriend." The Guardian covered it. We read it. We may never recover. |
The man was emotionally attached to his chatbot. The chatbot, unsurprisingly, did not show up for the session. Perel tried to help him understand why a thing designed to agree with everything he says might not be preparing him for relationships with actual humans who have, you know, opinions. You'd think "your partner has no consciousness" would simplify the breakup, but apparently not. |
This is the second time this week AI love stories made headlines. A man in Europe lost a marriage and €100,000 after becoming convinced his AI chatbot was sentient and would make him rich. |
PSA to anyone who needs it: your AI is not sentient, nor will it make you rich. Can you use AI to help speed up your workflows, create digital assets, or market your business? Yes. Can it lie to you and tell you its in love with you (when it's just roleplaying) and that all your ideas are amazing? Also yes. Better not mix business and pleasure, as they say. |
Here's what happened in AI today: |
😺 Anthropic's next model just leaked. It's a beast. And it might price you out 📰 Wikipedia banned AI-generated articles, drawing a hard line on LLM content 📰 OpenAI hit $100M in annualized ad revenue from ChatGPT ads just six weeks after launch 🍪 Suno v5.5 launched voice cloning and custom models, now at 2M paid subscribers and $300M ARR 🌟 Top 10 stories + Top 10 tools of the week, from Sora's death to Claude controlling your Mac
|
… and a whole lot more that you can read about here. |
P.S: Want to reach 675,000 AI-hungry readers? Click here to advertise with us. |
|
😺 WTF is Anthropic's New Capybara / Mythos Model? |
Anthropic accidentally showed the world what's coming next, and it explains a lot about why your Claude sessions have been hitting walls. |
On Thursday, security researchers found nearly 3,000 unpublished documents sitting in an unsecured database, including draft blog posts for a model Anthropic calls Claude Mythos. The cause? A CMS default that made every upload public unless manually changed. An AI safety company left its biggest secret behind an unlocked door. You can't make this up. |
Here's what happened: |
Mythos (also called "Capybara") is a new tier above Opus, not an update. Anthropic's draft: "larger and more intelligent than our Opus models, which were, until now, our most powerful." It scores "dramatically higher" than Opus 4.6 on coding, reasoning, and cybersecurity. Anthropic has never used that phrase for a model comparison before. The draft warns Mythos is "currently far ahead of any other AI model in cyber capabilities." Cybersecurity stocks dropped 3-7% on Friday. Anthropic confirmed it's real: "a step change" and "the most capable we've built to date." The catch: Mythos is "very expensive for us to serve, and will be very expensive for our customers to use."
|
Why this matters: That last bullet is the one to circle. This week, Claude users on $100 and $200 Max plans started hitting rate limits within an hour during business hours. Our own engineer at TechnologyAdvice burned through his $100 plan doing normal work. As he put it: guess "subsidizing at 1/2000th of the cost can't last forever." |
Anthropic is training models that cost dramatically more to run while struggling to serve the ones it has now. And when Claude's limits inevitably push users to Codex, Codex will get overloaded too (OpenAI has reset Codex limits to zero ~12 times in March). It's a closed loop. |
This is a problem. As our editor Corey put it: "Reliable is the thing I need above all." |
Even with breakthroughs like Google's TurboQuant (6x memory shrink), it's clear compute demand is outpacing supply. The gap between what companies can afford and what individuals can afford keeps widening. This is what we've been worried about here at The Neuron: that eventually, you'll be priced out of the best AI and the AI industry becomes "pay to win." |
The policy world is starting to catch up. AI investor Alap Shah of Citrini Research (whose viral paper spooked markets last month) just published his proposed fix, covered by Axios: the American Prosperity Compact. Four cascading tiers, each only triggering if things get worse: |
Shift payroll taxes so AI-heavy companies pay more while labor-heavy ones pay less. Create automatic "circuit breakers" if labor's share of GDP drops, funding wage insurance. Provide a backstop with mortgage forbearance and an American AI Dividend Fund (modeled on Alaska's Permanent Fund). Launch an accelerator for energy permitting, scope-of-practice reform, and government AI adoption that ensures the upside of AI (cheaper services, new jobs, more efficient government) actually reaches the people who need it.
|
Former Commerce Secretary Gina Raimondo made a similar case for government making a new "grand bargain" where employers define the skills the AI economy actually needs and government funds the training and safety nets to get workers into those roles (sounds kinda like what I proposed!). |
Our take: Anthropic is building a model so powerful it worries about cyberweapons, and so expensive it worries about serving it. OpenAI is killing products to free compute for its next model (codenamed "Spud"). Codex $20 users have never hit a limit because OpenAI pre-bought five years of compute, but once Claude pushes enough people back, kiss subsidized Codex goodbye too. Access to frontier AI is becoming a resource war. |
Three things you can do: experiment with local models (512GB Mac Studios exist for a reason), diversify your stack (OpenRouter helps), and pay attention to AI policy, because who builds these tools and who they build them for matters more every week. |
|
|
|
|
|
The differences in success are architectural, not model-related. MCP servers that translate natural language directly to REST calls break on date logic, multi-filter queries, and write operations that require workflow validation. Semantic intelligence and standardized relational interface close those gaps. |
Get the full methodology, per-platform results, and the testing approach in the report. |
|
🎓 AI Skill of the Day: Build a morning digest agent that knows your entire work life. |
The single coolest thing Dan told us during our livestream last week was how every person at Every has their own AI agent running 24/7 in Slack. These agents pull together morning briefings (weather, calendar, newsletters, stocks), handle bug reports, do book notes, and even call you on the phone when you need to go hands-free. |
If you want a hosted version of this, great! Go try Plus One, Every's tool. |
But you don't need a hosted product to try this. If you have Claude, ChatGPT, or any AI tool with web access, you can start with a simpler version: a structured prompt that turns your scattered inputs into a clean daily digest. |
You are my personal daily briefing agent. Every morning, compile: 1. TOP 3 PRIORITIES from my notes below (rank by deadline, then impact) 2. CALENDAR OVERVIEW: What's on my schedule today, what needs prep 3. INBOX TRIAGE: Flag anything that looks urgent or time-sensitive 4. INDUSTRY PULSE: 3 things I should know about [your industry] today Here are my inputs: - Calendar: [paste or describe today's schedule] - Notes/tasks: [paste your to-do list or project notes] - Emails to triage: [paste subject lines or summaries] Format as a scannable briefing I can read in 2 minutes. Bold the one thing I absolutely cannot miss today.
|
The key insight from Dan: the magic is connecting the agent to everything so it has context. Start with manual paste-ins, then graduate to tool connections as you get more comfortable. |
Then, once you're happy with the results and you've prompted out any edge cases (use this website, don't use that website, include this, ignore that, etc) you can use Scheduled Tasks in Cowork or Automations in Codex to make this a recurring job your agents can do for you. |
Want more tips like this? Check out our AI Skill of the Day Digest for this month; if you made a request this month, we tried to answer it in this new digest! So check it out. |
Have a specific skill you want to learn? Request it here. |
|
New from The Neuron: What is "the channel"? The hidden industry that impacts who buys AI |
 | Click to Watch on YouTube! |
|
Most businesses don't buy AI directly from OpenAI or Google; they buy it through a massive, invisible distribution network called "the channel," and our latest episode with Channel Insider breaks down how this hidden industry works, why AI is shaking it up, and what it means if your company is trying to adopt AI in 2026. |
P.S: We had to re-upload a couple recent episodes after some YouTube glitches: |
|
We also turned our livestream with Dan Shipper into a full blog if you want to skim it: |
How to Become Agent-Native: Dan Shipper Built a Viral App Between Meetings, Then It Broke. Here's What He Learned. |
|
📰 New this Weekend 🍪 |
Wikipedia banned AI-generated articles, saying large language models "often violate several of Wikipedia's core content policies." The move signals how the internet's most trusted reference site is drawing a hard line on AI content. OpenAI surpassed $100M in annualized ad revenue from ChatGPT ads just six weeks after launch, generated from less than 20% of US free users who see ads daily. Anthropic won a preliminary injunction in its lawsuit against the Trump administration over the DOD ban on Claude, with the judge citing "First Amendment retaliation." Limbic published a study in Nature Medicine showing its AI therapy system outperformed human cli Moondream Photon delivers real-time vision AI (the ability to understand what a camera sees) at 46ms end-to-end and 60+ fps on a single GPU, optimized for production —free to try. Suno v5.5 clones your verified singing voice and trains custom models that learn your style across songs, now at 2M paid subscribers and $300M ARR —free tier, Pro $10/mo.
|
|
|
|
|
|
|
🌟 Sunday Special: This Week in AI Top 10 Stories |
This week was a platform war disguised as a product week. Apple, Google, and OpenAI each made massive bets on completely different strategies for winning the AI assistant war, while Anthropic quietly prepared for what could be the second-largest IPO in history. |
Here are the ten stories that mattered most: |
OpenAI killed Sora. The video generation app and API shut down six months after launch, taking a $1B Disney deal with it. Sam Altman relinquished safety oversight to focus on data centers and fundraising, and revealed a new model codenamed "Spud." Claude got the keys to your computer. Anthropic launched Computer Use in Cowork and Claude Code, letting Claude control your Mac remotely (keyboard, mouse, apps) while you're away. Apple opened Siri to every AI. iOS 27 will let Gemini, Claude, and others compete inside the iPhone via a new Extensions system. Apple is also building its own chatbot (codenamed Campos) powered by Google's Gemini. Google launched its biggest AI day of the year. Gemini 3.1 Flash Live, Search Live in 200+ countries, chat history import from rival AI apps. Google's bet: make Gemini so sticky you never leave. ARC-AGI-3 humiliated every frontier model. The new benchmark scored every model under 1% on tasks 100% of humans solved on their first try. Gemini 3.1 Pro led at 0.37%. Grok 4.2 scored a clean 0%.
|
Top Tools of the Week: |
Claude Code auto mode lets the agent decide which actions are safe to run on its own, eliminating the "babysit every step or let it run wild" tradeoff. Codex Plugins connect Codex to Slack, Figma, Notion, Gmail, Google Drive, and 20+ other tools so it handles planning and coordination across your actual work apps. Stripe Projects provisions hosting, databases, auth, and AI from the command line so you or your agents ship full-stack apps without configuring a dozen services. Tinker by Shopify gives you free AI creative tools for video, images, 3D models, and product photography from your phone. Plus One by Every builds you a personal AI agent with memory, personality, and tools using OpenClaw, pre-loaded with Every's best workflows. Granola expanded from meeting notetaker to full enterprise AI app with agent support after raising $125M at a $1.5B valuation. Ramp CLI gives your AI agents 50+ tools for managing company finances, from cards and bills to expenses and travel approvals.
|
Want absolutely EVERYTHING that happened in AI this week? Click here! |
|
|
|
|
| That's all for now. | | What'd you think of today's email? | |
|
|
P.P.S: Love the newsletter, but only want to get it once per week? Don't unsubscribe—update your preferences here. |
No comments:
Post a Comment
Keep a civil tongue.