Welcome, humans.
So, the India AI Impact Summit wrapped up this week in New Delhi, and it was… a lot. Google DeepMind's Demis Hassabis said AGI could arrive within five years (half his previous estimate). Sam Altman said superintelligence is "a couple of years away" (that link is a great interview btw). And Anthropic's Dario Amodei predicted AI could deliver 25% annual GDP growth for India. Bold claims from three guys who wouldn't even hold each other's hands on stage.
I know we’ve brought this up like three times now, but we’re still not over it. It’s the most hilariously awkward meme we’ve seen all year. And ALMOST as cringe as Dario’s “reading off his phone” speech.
No shade, as someone who says a lot of “ums” and “you knows” and “like”s in my own public speaking, I get it. We’re about to deep dive on Claude Code tho so we thought it only fair to call him out. Y’know, in the spirit of “true equilibrium.”
Here’s what happened in AI today:
70+ countries signed the "Delhi Declaration" at India's AI Impact Summit.
Amazon said hackers used AI to breach 600+ firewalls across 55 countries.
OpenAI scaled back its compute spending target to $600B by 2030.
A Cambridge study found only 4 of 30 top AI agents published formal safety evals.
Click to watch on YouTube!
Do you work in customer service? Or use customer service in your business? Then you’re going to want to watch our latest podcast episode with Matt Price of Crescendo AI on the future of customer service in the AI age.
Crescendo’s approach is unique; instead of fully replacing humans, they’re betting on turning call centers into a place where humans manage agents, using AI and humans together to improve customer experiences…and anyone can use their system to do the same.
After coding, customer service was predicted as the next area to get fully automated… so where are we actually at in that cycle? How’s that going? Well, Matt has some unique takes on where we’re at and where the industry goes from here.
Watch and/or Listen: Spotify | Apple Podcasts | YouTube (click above!)
Your Computer Can Now Do Your Monday Morning Busywork While You Sleep
You know that feeling on Friday morning when you realize you forgot to reply to three urgent emails from Tuesday? Yeah, same.
Well, AI educator Allie K. Miller got so fed up with this that she built a system inside Claude Code (Anthropic's command-line tool) that automatically scans her Gmail every Friday at 9 AM, finds every urgent email she hasn't responded to, and sends herself a summary with direct links to each thread.
Here's the gist of how it works:
Custom commands are reusable instructions saved in Claude Code. Miller's /ue command checks her inbox (read AND unread), flags urgent messages, and checks if she's already replied.
MCP connections let Claude Code talk to your Gmail, Slack, Notion, or calendar. Think of it as giving Claude a set of keys to your work apps.
Cron jobs run your commands on a schedule automatically; every morning, every Friday, whatever you want.
Stacked commands are the power move. Miller built a daily briefing that checks her calendar, then automatically runs a separate client-prep command for every meeting that day. She wakes up to fully compiled briefings.
The whole setup takes 2-3 hours the first time (mostly getting Google permissions sorted). Miller says it saves that much time in a week or two.
Is all that too technical for you? Don't want to mess with cron jobs? Well, runCLAUDErun is a free Mac app that lets you schedule Claude Code tasks with a visual interface instead of terminal commands. Same idea, friendlier packaging.
There’s also Tasklet, which abstracts away the horror of the technical setup of all this stuff, and just does it for you. But if you want to set it up yourself with the most amount of control, you can follow Allie’s tutorial and empower yourself to learn how to automate not just your emails, but basically anything.
Why this matters: The “what should I actually use AI for?” question finally has a concrete answer for most knowledge workers. You're not replacing your job; you're automating the 30 minutes of information-gathering busywork that sits between you and the work that actually requires your brain.
And once you set it up once, it runs forever. Well, as long as your laptop is open; baby steps! Now, for something that operates 24/7 and is a bit more complicated, you can set up OpenClaw… and even use it to start making money on Upwork right now.
FROM OUR PARTNERS
It’s 2 AM. A Customer Just Called. XBert Already Answered
Most businesses lose leads to voicemail. Yours won’t. XBert is the AI receptionist that handles calls, texts, and chats 24/7. It books appointments, answers questions, captures every lead, and routes urgent requests to your team instantly. While your competitors let calls ring out, your business never misses a beat. Trusted by 100K+ businesses with a 4.9/5 rating. Set up in minutes and start your free 7-day trial today.
Prompt Tip of the Day
Stanford AI computer scientist Harper Carroll has a great prompt tip for you today:
If you’re using a non-reasoning AI model (so the faster / free ones), you should put your task (the question you want the AI to answer) BEFORE the context (any text or descriptions or other relevant content) in your prompt.
This causes the AI to read the context you give it with the task to be done front and center.
She says this advice actually works for human comprehension, too: during the reading comprehension portion of her SATs, she was told to read the question to answer before reading the essay, and it worked (she got into Stanford, after all).
There’s some free SAT advice for any parents out there!
Want more tips like this? Check out our Prompt Tip of the Day Digest for February.
Treats to Try
*Asterisk = from our partners (only the first one!). Advertise to 650K+ readers here!
*Outskill is hosting a 2 day LIVE AI Mastermind this weekend between 10am-7pm EST where you'll build automations, create personalized agents, and learn to turn AI into your ultimate competitive edge. Register here before they run out of seats (free for next 72 hours only).
Anthropic launched Claude Code Security, which scans codebases for vulnerabilities and auto-suggests patche
Decagon builds customer service agents that handle your support tickets across chat, email, voice, and SMS—upgrading flights, reordering cards, processing returns—24/7 without escalating to humans (raised $250M).
Recare's agent handles patient discharge paperwork—processing PDFs and clinical notes, creating transfer letters, finding available care spots—replacing faxing and phone calls for hospitals (raised €37M).
Mozart AI lets you go from an idea to a finished song, mastered track, and music video in minutes; just launched on iOS so you can create on the go (raised $6M).
Reload launched EPIC, a Cursor extension that gives your coding agents persistent memory of your project's architecture, so they stop losing context between sessions and drifting from the original plan—free to try, then $20/month.
NVIDIA open-sourced DreamDojo, a robotics world model trained on 44K hours of human video that predicts physical interactions without a physics engine.
Around the Horn
Whoever made this: Why? 🤣 Also, remember when viral AI content was just this kinda stuff and not, idk, swarms of agents starting their own religion? P.S: there’s also a Snoop Dogg version. The names rule.
OpenAI scaled back its compute spending target from $1.4 trillion to $600 billion by 2030, while projecting $280 billion in revenue.
The India AI Impact Summit's "Delhi Declaration" was signed by 70+ countries, but the White House flatly rejected global AI governance; official Michael Kratsios said, "We totally reject global governance of AI."
Amazon says a small group of Russian-speaking hackers used commercial AI tools to breach 600+ Fortinet firewalls across 55 countries in weeks, exploiting weak passwords and exposed ports at a scale Amazon said would have been impossible without AI.
A Cambridge-led study found only 4 of 30 top AI agents published formal safety evaluations, with browser agents missing 64% of safety disclosures despite being the most autonomous.
NEW: Want more? Check out our new Around the Horn Digest for February here
FROM OUR PARTNERS
Stop typing prompt essays. Dictate full context and Wispr Flow turns it into a clean, structured prompt you can paste into your AI assistant in seconds.
Sunday Special
This means Opus 4.6 can handle 14.5-hour software tasks, but only gets them right half the time. But if you need 80% reliability, it tops out at 1-hour long tasks.
Welcome to vibe era of AI, says AI Explained (love these videos).
Dr. Hannah Fry explains that treating AI as an infallible, god-like entity rather than a highly capable tool is already causing real-world harm.
Nick Sareev shows how to one shot a beautiful website with Gemini (prompt).
A Cat’s Commentary
 | That’s all for now.
|
