I want to be useful today rather than philosophical. Useful in the sense of: here are specific tools, here is what I actually use them for, here is what they're good at and where they fall short.
There are a lot of AI tool lists online. Most of them are either affiliate-link parades or slightly outdated by the time they're published. I'm going to give you the five I genuinely reach for every day in the operation, with enough context that you can evaluate whether they'd be useful for what you're doing.
1. Claude (Anthropic)
This is my primary writing and reasoning tool. I use it for long-form drafting, editorial feedback on my own writing, research synthesis, and thinking through complex problems out loud.
The thing Claude does that I find most valuable: it will push back. If I give it an argument and it finds a flaw, it'll tell me. If I ask it to review a piece I've written and there are structural problems, it doesn't just validate me. For someone who spent fifteen years in an adversarial legal environment where assumptions got stress-tested as a matter of course, a tool that maintains intellectual friction rather than giving me what I want to hear is genuinely useful.
Where it falls short: it's slower than some alternatives on quick, structured tasks. If I need something simple done fast, there are better options. But for anything that requires sustained reasoning or genuine writing quality, it's where I spend the most time.
2. ChatGPT (OpenAI)
I use ChatGPT primarily for structured, defined tasks — building outlines from a set of inputs, producing first-draft copy for defined purposes, analysis of structured data. It's faster than Claude for many tasks and excellent at following explicit, detailed instructions.
I also use it as a sounding board for ideas when I want quick iteration — bouncing options back and forth at a pace that doesn't require me to write out everything formally. For rapid ideation, it's excellent.
The two tools complement each other more than they compete. I'll often draft something with ChatGPT and then take it to Claude for deeper revision.
3. Midjourney
For visual content — featured images for newsletter issues, directory imagery, product design starting points — Midjourney is where I go. The quality ceiling for photorealistic and stylized illustration is higher than any other tool I've used, and the control over aesthetic is good enough that I can produce something that looks consistent with a brand identity rather than just "AI-generated."
The learning curve is real. The prompting language for Midjourney is its own skill, and it took me several weeks of consistent use before I could reliably produce what I was trying to produce. But the investment pays off. I have not hired a designer for the newsletter or the directories, which is a real cost savings.
4. Perplexity
Perplexity is how I do research. It's a search engine that uses AI to synthesize results rather than giving you a list of links to click through. The combination of web-current information with synthesized summaries means I can get oriented on a topic faster than any other research workflow I've used.
The important caveat: I verify anything meaningful. Perplexity is a starting point, not an endpoint. But for understanding a space quickly, checking whether a market exists, or finding relevant context for a piece I'm writing, it's cut my research time significantly.
5. Beehiiv's built-in tools (and the Beehiiv platform more broadly)
This one is a little meta, but it belongs on the list. Beehiiv has been rolling out AI-assisted features for newsletter production — subject line testing, analytics on what content performs, subscriber segmentation. The platform itself is part of the daily toolset, and the quality of its analytics infrastructure makes it genuinely useful beyond just being a publishing mechanism.
What's not on the list and why.
Several tools that get a lot of attention are not in my daily rotation. Jasper and Copy.ai are capable tools but I find the primary models — Claude and ChatGPT — more flexible and ultimately more capable for what I'm doing. Various automation platforms are useful but are not things I interact with daily; they run in the background. I'll cover the full automation stack in a future issue.
The honest meta-point.
The question of which tools to use matters less than how you use them. Five people could have identical tool stacks and produce very different results depending on their prompting skill, their editorial judgment, and their clarity about what they're trying to produce. The tools are multiplicative — they amplify whatever judgment and skill you bring to them.
That said: using the right tool for the right task instead of forcing one tool to do everything is real. Like any craft, the right instrument for the job matters.
This Week in AI: The competition between AI labs has driven model release cadences to a pace that makes it difficult to keep a tool recommendation current for more than a few months. The practical advice: evaluate tools based on your specific use cases, not on benchmark comparisons, and be willing to revisit your stack every quarter or two. The hierarchy shifts.
The full toolkit — including our complete AI stack with pricing and recommended use cases for each tool — is free at start.tenstreamslab.com. Grab it.