If you tried AI a while back and decided it wasn’t worth it, your opinion is running on old information. Here are the five updates that matter most, what each one is, and how to use it.
If you think AI is overrated, there’s a good chance your opinion is just out of date.
Most people who say it isn’t useful tried it once, got a bland answer, and went back to the old way. That’s not what exists now. For each update below I show you what it is and how to use it today.
| What’s new | Why it matters | How to use it |
|---|---|---|
| Agents do real work | They finish whole jobs instead of answering questions | Hand Claude or ChatGPT Work a job and a folder |
| Images are good enough to ship | Ad-quality visuals from a text box | Describe the finished shot in Gemini or ChatGPT Images |
| Talking beats typing | You give far more context out loud | Use Wispr Flow or voice mode |
| Desktop beats the browser | The AI can work with your real files | Connect one folder in the desktop app |
| Intent beats instructions | Describing the goal beats scripting steps | Say what done looks like and give context |
This is the biggest one. When you tried AI, you asked a question, got an answer, and still did the work. Now you can hand it a whole job and walk away while it runs.
01
Where the agents live
The friendly versions are Claude and ChatGPT Work. On Claude’s Pro, Max and Ultra plans, Cowork is being merged into regular chat, so you just describe the task in any conversation and it hands back finished spreadsheets, slides and docs. ChatGPT Work lives in the new ChatGPT desktop app, next to Chat and Codex.
One level deeper are Claude Code and Codex. They started as coding tools and became some of the most capable general agents out there. They work straight from the files on your computer. You give the outcome, they figure out the middle.
I run one agent for my brand’s email marketing and another for my whole content pipeline. That only works because I stopped asking questions and started giving jobs.
Want the longer version? Here’s how to build your first AI agent and how Cowork works.
Remember warped hands and text that came out as alien letters? That era is over. I use AI images in my videos, paid ads, presentations and product mockups. Where I used to open Canva, I now start here.
02
Pick the right one
Nano Banana in Gemini when it has to look real or hold correct text, like product shots, posters and infographics. It’s also great at blending your own reference photos into one scene. Nano Banana 2 is the default, and if you pay for a Google AI plan you can open the More menu under an image and pick Redo with Pro.
ChatGPT Images 2.5 when you need it to follow exact directions and edit patiently. Think logos, clean layouts, mockups, and changing one small thing without the whole picture shifting.
03
The tricks that matter
Describe the finished image like a creative director. "Dog, park, 4k" gives you generic soup. Name the subject, setting, framing, light and mood.
Say what you want, not what you don’t. "An empty street" works better than "a street without cars."
For text, put the exact words in quotes and say they should appear once and nowhere else.
Give each reference photo a job. "Use photo one for the product, photo two for the background, keep the product identical."
Change one thing at a time, and add "natural skin texture, real photo feel" to kill the plastic AI sheen.
More on the ChatGPT side in my ChatGPT Images guide.
The simplest upgrade I ever made: I stopped typing to AI and started talking to it. I use Wispr Flow all day. It works anywhere you can type, including Claude, ChatGPT, email and notes, and turns your talking into clean text with the filler words gone. It runs on Mac, Windows, iPhone and Android, and the free plan includes a weekly word allowance, which is smaller on iPhone. Full walkthrough in my Wispr Flow guide.
Rather not add a tool? Voice mode is built into ChatGPT, and Claude’s voice mode is on every plan, including Free.
Here’s why it matters. When you type, you shrink what you meant. You cut the backstory and the constraints, then blame the AI for a generic answer. When you talk, the detail comes out, and detail is what makes the answer good.
Most people still use AI in a browser tab, where it only knows what you paste in. So you spend the day copying things out of documents and into a chat box.
04
The AI can reach your actual files
In the Claude desktop app, you connect a folder and Claude can read, edit and organize those files. It only reaches the folders you connected, and only while the app is open. This part needs a paid plan. (Claude help)
The ChatGPT desktop app for Mac and Windows puts Chat, Work and Codex in one window, and Work and Codex can use files on your computer. Both apps are free to download.
The task that used to be impossible because it involved eleven documents is now a normal request.
This one isn’t a feature, which is why almost nobody noticed. You don’t need to tell AI exactly how to think anymore. Older models needed the steps spelled out. Today’s models do better when you spend your effort on the goal and the context.
05
Context and intent beat instructions
Old way. "Read this transcript, pull the key points, group them into themes, summarize each, then add an intro and a conclusion."
New way. "Turn this transcript into a one-page brief my team can read in three minutes and know what we decided and who owns what."
The second one is shorter and better, because it tells the model what the thing is for.