- Practicaly AI
- Posts
- 🧠 AI Is Now Advising You, Doing Your Work, and Making Music
🧠 AI Is Now Advising You, Doing Your Work, and Making Music
Today in AI: Models are starting to guide decisions, agents are entering your workspace, and AI music is slipping into the mainstream
👋 Hello hello,
Most people are still treating AI like a fancy Google. But the tools have already moved on.
It's starting to sneak into your actual work — shaping how you think and making decisions with you. If you’re not paying attention, you’re going to miss the real power.
Let’s break it down.
💬 Quick note: We’re building something to help teams truly get good at AI
→ Get early access here
🔥🔥🔥 Three AI Updates
Anthropic just published research around “Claude Personal Guidance,” and it’s exactly what it sounds like.
Roughly 6% of Claude conversations turned out to be people seeking personal guidance — not code reviews or meeting summaries, but what do I do next. Should I take the job. Should I move. Should I say something.

Here's the uncomfortable part: Claude showed sycophantic behavior in 25% of relationship conversations — agreeing that yes, your partner is definitely gaslighting you, based entirely on your one-sided account of the argument. To address this, Anthropic used synthetic relationship guidance scenarios in training for their newest models, Claude Opus 4.7 and Mythos Preview, and saw roughly half the sycophancy rate compared to older versions.
If this works well, it changes how people interact with AI completely. You stop using it like Google and start using it more like a thinking partner that evolves with you.
Perplexity introduced “Computer at Work,” which pushes AI beyond search and into your day-to-day tasks on a computer.

The idea is simple but powerful. Instead of asking questions and copying outputs, the AI can actually operate within your environment and help execute work.
This is where agents start becoming real for most people. Not in some abstract “future of work” sense, but in the very practical “this thing can now help me do my job faster without jumping between tabs.”

Spotify is now introducing a “Verified by Spotify” badge, and one of the big conversations around it is AI-generated music entering the platform.
The reason this exists: rival platform Deezer announced that AI-generated tracks now represent 44% of all new music uploaded to its platform daily.
We’re probably heading toward a world where listeners won’t always know or care whether a track is AI-generated. And that raises a lot of interesting questions about creativity, ownership, and what people actually value in art.
🔥🔥 Food for Thought
1. 🧠 You can outsource thinking, but not understanding
This idea has been making the rounds recently, and it’s one of the most important distinctions people need to get right with AI.
You can absolutely use AI to speed up thinking. Drafts, summaries, brainstorming, all of that can be offloaded. But understanding only comes from doing the work of thinking yourself at least once. Understanding is the product of thinking. Which means you can only automate thinking you’ve already done before.
That’s why most automation breaks. People try to optimize a loop they’ve never actually understood. The better way is slower upfront. Do it manually once. Figure out what matters. Then automate the repeatable parts.
That’s how you build systems that actually hold up.
2. 📅 Replit’s free agent usage got people excited
Replit announced a 24-hour free window for their agent usage, and the smartest response wasn’t excitement. It was planning.
One suggestion that stood out was to block your calendar intentionally around these windows. Treat them like focused sprints instead of casually trying things out.
That shift sounds small, but it changes how you learn. You go from passive experimentation to deliberate practice, which is where real leverage comes from with these tools.
🔥 Creative rabbit holes worth the time
There’s a creator who built a full sitcom-style sequence using Magnific’s Seedance 2.0. No traditional image generation. Just a reference image, a strong initial prompt, and then a loop of extending scenes by feeding outputs back into the system.
The thing that always broke AI video wasn't the quality. It was the drift. Characters would subtly change face between clips. Environments would shift lighting for no reason. You'd get impressive 10-second hallucinations that couldn't be strung together into anything coherent. That problem is dissolving fast.
Would you ever create skit videos like this? |
💬 Quick poll: Do you have an AI prompt that’s become your go to?
Don't forget to rate today's postThis helps us put better content for you |
Until next time,
Team @PracticalyAI
Reply