I’m a new DEVONthink user and am getting pretty impressed by the software coming from DT. I’m wondering, though, if there remain good reasons to invest in DEVONagent when standard search engines (and souped up browsers like Arc) are AI-enhanced now.
Can anyone persuade me to spend a little more of my kids’ inheritance?
Have you tried DEVONagent yet?
with DevonAgent no Cookies will be installed so Google cannot trace you - in the browser it can
How long is a piece of string?
That is, whether DevonAgent is useful depends on what you’re actually trying to do. The search functions, IMO, are only a gateway to what it can do with the results.
In addition to the cookie thing, remember that Google and Bing exist to feed you advertising. That is, their financial success is only tangentially connected to the accuracy of their results.
Have you tried DEVONagent yet?
Not yet, but I will. I just completed my trial of DEVONthink Pro and have joined the cult many enthusiasts. It’s such an impressive piece of software.
That is a very convincing point. Thanks.
Very good point, @Ja_acov. And maybe there’s something to be said for a piece of software that has intelligent search as its primary purpose. I used to believe that about Google, but that was when it first appeared back in the 1900s.
Back when their company motto was “Don’t be evil.”, as sentiment long lost on them.
I was just talking to our team about that very thing this morning 
I often find myself telling people that they’ve removed the word “Don’t”. 
I deal with other teams who receive unpredictable occassional updates to a publicly available website. They have an informational advantage over me in that they are sent the new information and I am not. I am too lazy to check the website for updates each time I engage with these other teams. So, DEVONagent Pro is a useful tool to scan the websites for updates and email them to me.
Compared to the ai search results I see on bing and google, Devonagent actually returns results. Actual results relevant to what I’m seeking.
Better yet, I can hand those results off into all kinds of places and formats. For instance, I can have a local model synthesize and process the info it found. And I have an auditable trail of information leading right back to archived source materials.
I recently came across a perspective on AI-based search vs. more traditional research methods, and it resonated with me—especially in the context of tools like DEVONthink.
Imagine you’re researching a topic like 18th-century Chinese porcelain.
With AI, you ask a question and get a clear, direct answer. It’s efficient—but also quite narrow.
Now compare that to a more traditional approach:
searching with Google (or Kagi in my case), opening multiple articles, following links, or even going to a library and browsing books or papers. It’s slower, no doubt. But along the way, you inevitably encounter adjacent ideas—unexpected stories, historical context, technical details, economic implications, influences on other crafts…
In other words, you don’t just get an answer—you build a richer mental map of the subject.
An analogy that came to mind:
Using AI is like Frodo taking a jet straight to Mordor and dropping the ring in. Mission accomplished—fast.
But we miss everything that made the journey so fun!
Traditional (re)search, on the other hand, feels like actually going through Middle-earth. It takes time, but the depth, the connections, and the experience are what make it valuable.
On a more personal note, I’ve been thinking about how much friction we’re removing from everyday life. While it brings comfort, it can also strip away depth and learning. Cooking is a simple example: preparing a meal yourself versus ordering delivery. One is faster—but the other builds skill, understanding, and a relationship with the craft.
I think Frodo would say you’ve got a weird definition of fun. If it had been possible to have one of the Eagles drop it in the fire that would have been much more enjoyable all around.
But to your point, yes, exactly. This is why writers compare “writing” with AI to “running” a marathon by bicycle.
Except that the bicycle probably will follow the course fairly reliably, while AI might give an answer that is clear, direct, and wrong. And unless you’ve got a mental map, you won’t necessarily know.
“Clear, direct, and wrong” — honestly sounds like a perfect slogan for an AI startup in some parallel universe 
I’ve honestly lost count of how many times ChatGPT has given me a wrong answer — with absolute, confidence 