I gave Claude Cowork 7 non-coding jobs, and it earned a spot in my toolbox


7 powerful ways Claude Cowork helped me beyond writing code

Elyse Betters Picaro / ZDNET

Follow ZDNET: Add us as a preferred source on Google.


ZDNET’s key takeaways

  • Claude Cowork turned messy tasks into useful results.
  • It helped fix servers, sort PDFs, and review contracts.
  • Giving an AI that much access remains unsettling.

I am not a naturally trusting person. So when Claude Cowork was launched, the idea of giving an AI access to my Google Docs and Gmail did not sit well with me.

Heck, back in the day, the idea of letting Google manage my email didn’t sit well with me, either. I was once a big proponent of only using servers you could touch, restart, and disassemble at 3 a.m. The idea of letting Google, the self-acknowledged master of information vacuuming, have access to my email seemed ludicrous.

Heck, even letting a cloud hosting provider host my web server seemed extreme. There was a time in the early days of the web when my startup’s $2,000-a-month T-1 line entered my apartment in my bedroom, ran through my bathroom, then through my bedroom closet, across the hall, and into a linen closet with an octet of tower servers and an ambient temperature in the mid-90s.

Try explaining that to your apartment manager. And no, it wasn’t a grow room, which I had to prove to multiple authorities over the years. My only saving grace was I could fix their computers for them.

Also: I tested the new Claude Desktop on Linux – here’s how it compares to rival apps

The point is that I’m a bit of a control freak. Letting Claude Cowork loose to do work for me triggered my control freak alarms in a big way. But it’s been slowly worming its way into my life the same way cloud hosting providers and Gmail did: it’s able to save me time.

I’m still not comfortable with it, but over the past few months, I’ve set it loose on some projects where it came through, saving me hours of tedious work. In this article, I’m going to give you a rapid-fire summary of seven of those projects.

If you want more information, I’ve already written detailed articles about a few of them. I’ll provide links for you, so you can dig in. Also, please note that I’m using Cowork on the $100-a-month Claude Max plan. Cowork will also work on the less expensive $20-a-month Claude Pro plan, but you’ll run out of usage allocation faster.

1. Analyze my Home Depot spending

The very first thing I tried, back in January when Cowork was just a shiny new beta release, was to have it dig through my Home Depot statements and try to come up with an analysis of what I bought. I fed it a cluster of Home Depot statement PDFs and set it loose. I felt safe-ish doing this because the PDFs didn’t show my full account number.

Also: Claude Fable 5 is back, but I’m sticking with Opus 4.8 for daily work: 5 reasons why

This was only marginally successful, but it wasn’t Cowork’s fault. The Home Depot statements only included general categories and no full item numbers, so I wasn’t easily able to tell how much I spent on tools versus plywood, for example.

2. Organize my PDFs

The second thing I did, as documented in the same article I just mentioned, was to have Cowork dig through a copy of my Downloads folder and organize my PDFs.

Also: I compared Claude Opus 4.8 with 4.7 in a 10-round honesty test – and a legal prompt broke it

This took things a level above Hazel, which is a Mac tool I use to sort files into folders by file type and year. I wanted Cowork to examine the PDFs and divide them into categories.

As you can see, it created a solid taxonomy based on the PDFs themselves. It also discovered that many of the PDFs started out with random letters and numbers for names, so it examined them and renamed them to be clearer.

I still haven’t been brave enough to let this loose in my actual Downloads folder, but I’m reaching the point where it’s so good, I just might do that one day. I did like how it categorized everything.

img-9875

David Gewirtz/ZDNET

3. Lawyer in a box

Back in May, Anthropic introduced its skill set for small businesses. Within that package was an amazingly powerful skill that runs inside Claude Cowork. Once you install it, you trigger it by typing /review-contract.

Also: Why I ditched Copilot for Claude in Word, Excel, and PowerPoint – and how you can, too

Feed it a contract in Word or PDF form, or as any other doc format Claude reads, and Cowork does a very detailed, very complete, and almost jaw-droppingly excellent analysis of the contract. I ran a few of my old contracts through it, and Cowork found some items I had been worried about and some items that had never surfaced during the original review.

My take on this one is that if you ever have a contract, use this.

4. Fixing a stuck server

Back in March, I told you about Karakeep, a tool for organizing articles and YouTube videos. It’s my locally hosted homelab version of Pocket. For months, it worked just fine. Then it stopped working. The thumbnails that are a signature feature of Karakeep just stopped being generated.

The architecture of this is typical homelab complexity. It’s a Docker installation, managed by Portainer, running on Linux. Anything could go wrong. Finding it could take days, and my sanity.

Also: I tried a Claude Code rival that’s local, open source, and completely free – how it went

So I set Cowork loose on it. I let it use my browser. I let it connect to Portainer’s web interface. I even let it use Webmin to dig around my Linux server.

I started the process at 10:38 a.m. Cowork finished at 11:58 a.m. During that time, the AI and I talked back and forth, working out the issue. I can’t be sure how long it would have taken me on my own. The fact was, I wasn’t on my own. And before noon, a tool I’ve come to rely on was back up and running. Oh, and the problem was that the Docker user-defined network had no upstream DNS server that it could reliably count on. We fixed that.

5. Aggregating blood pressure readings

I’ve been very intent about optimizing my health for the last few years. In particular, I’ve been carefully watching what I eat and tracking my blood pressure. A few months ago, I had a doctor’s checkup. I wanted to bring him the previous two months of blood pressure history.

Also: I let Claude AI control my Mac, and it worked flawlessly – with only two minor issues

While I like using my Withings cuff because it records all the readings, exporting those readings is quite difficult. By the time I decided to pull the data, I couldn’t find a reliable way to export just the data I needed. My appointment was coming up very quickly. So I just opened the web app, pointed Cowork at it, and told it to dig through it all and build a spreadsheet.

It did. I quickly had a chart to present to my doctor. I didn’t even have to spend the whole morning cutting and pasting. Good news: My blood pressure is doing great.

6. Sifting through thousands of Gmail messages

This is the point at which my trust stretched more than I felt comfortable with, but it was worth it. As I wrote in a recent article, I get a ton of emails from folks who want me to write about them, or who want to talk to me about articles I’ve written. One week, there was some particularly hot news. I wanted to aggregate some of the comments I had been sent into a sentiment analysis article.

Also: I used Claude Code to vibe code a Mac app in 8 hours, but it was more work than magic

So, I let Cowork loose in my Gmail. It dug through the more than 7,000 emails I received that week. It found eight that met my criteria. It wasn’t just a search on a specific set of strings. Cowork had to read the content of each message for context and decide if it was a fit. I ended up with eight messages I could use in the article. This was a very powerful example of using Cowork as a research assistant.

7. Mitigating a spam attack on my server

Of all the Cowork projects I’ve tried, this is the one that gave me that “we’re living in the future” feeling. The quick backstory is that my server was under attack from spammers polluting the site and clogging the server database. My hosting provider reached the point where I was informed that if I didn’t stop it, it would shut me down.

Also: How to learn Claude Code for free with Anthropic’s AI courses – one took me just 20 minutes

I had already done some mitigation work with a WordPress plugin that I was working on with OpenAI’s Codex agent. Codex is a direct competitor to Claude Code. I use both, but I keep them each on their own projects. The WordPress plugin is a Codex project.

But my low-end plan had a fairly limited token allocation for Codex, so I divided up the work. Codex did the coding. Cowork did the analysis and mitigation strategy. This was a brilliant, almost nonstop team effort over a weekend. By Sunday night, the three of us were able to deploy a fix that stopped the spammers in their tracks. We also undid all their damage.

Cowork won a spot in my toolbox

I am still coming to terms with what Cowork can do, and deciding how much leash I’m willing to give it. Cowork also now has a mobile and web incarnation, which means it can be left to run without sitting on an open laptop. I haven’t tried that yet, but I regularly have work that could benefit from some help. I’ll probably delegate something to the cloud version of Cowork soon enough.

Also: I tested ChatGPT vs. Claude to see which is better – and if it’s worth switching

Also, note that none of these examples were really coding. Yes, the last example was coding-adjacent, in that it was advising the coding engine in Codex about strategy. But as Anthropic noted in its survey, Cowork seems to be for a lot more than code.

I am fairly convinced I’ll give this thing more and more to do. So far, the work has involved projects with a definable beginning and end. But with the web version, I may set Cowork to work on a schedule. I might have it do work that might have previously been given to something like OpenClaw, or that I just might have tediously and grudgingly done on my own by hand.

Stay tuned. As scary a tool as this thing is, it’s proven itself in its first six months to be enormously helpful. I look forward to further helpfulness in future months. That is, of course, if it doesn’t murder us all in our beds.

Would you use Claude Cowork on your files, your inbox, or your server if it could save you hours of work? Let us know in the comments below.


You can follow my day-to-day project updates on social media. Be sure to subscribe to my weekly update newsletter, and follow me on Twitter/X at @DavidGewirtz, on Facebook at Facebook.com/DavidGewirtz, on Instagram at Instagram.com/DavidGewirtz, on Bluesky at @DavidGewirtz.com, and on YouTube at YouTube.com/DavidGewirtzTV.





Source link

Leave a Reply

Subscribe to Our Newsletter

Get our latest articles delivered straight to your inbox. No spam, we promise.

Recent Reviews


TL;DR

Meta stripped NameTag facial recognition code from its AI app one day after WIRED exposed it on 50 million phones. Meta says no decision has been made.

Meta removed nearly all traces of an unreleased facial recognition system from its smart glasses companion app on Friday, one day after WIRED reported that the software had been quietly embedded in an app installed on more than 50 million phones. The feature, which Meta internally called NameTag, was designed to convert faces captured by the company’s Ray-Ban smart glasses into unique biometric signatures and compare them against a database stored on the user’s device. WIRED also found that faces the system failed to recognise were cropped, indexed, and stored locally for future processing.

Andy Stone, Meta’s vice president of communications, told WIRED on Monday that the feature is “purely exploratory,” adding that no final decision has been made on what to do with it. That characterisation sits uneasily with the evidence WIRED documented. The version of Meta AI published the day of WIRED’s Thursday report contained several code libraries explicitly named for face recognition, a process for running the NameTag recognition pipeline, and a “Person recognised” alert the app would have shown if someone were identified.

Friday’s release stripped all of it out, along with a folder where the app would have stored the cropped images and biometric signatures of unrecognised faces. Meta did not answer WIRED’s questions about why the code was removed or whether the changes were planned before the story was published. A few fragments remain in the latest version, including an internal debug menu label and a dormant link meant to open a recognised person’s profile, pointing to parts of the system that are no longer there.

The 💜 of EU tech

The latest rumblings from the EU tech scene, a story from our wise ol’ founder Boris, and some questionable AI art. It’s free, every week, in your inbox. Sign up now!

The gap between Meta’s public statements and the code WIRED found is the central tension. Before the Thursday report, Stone dismissed the findings by writing that the company could not answer questions about how the system would work because “the feature does not exist.” Andrew Bosworth, Meta’s chief technology officer, called the reporting “incredibly misleading” and “absolutely dishonest.” Yet the code was functional enough to include three AI models, one to detect faces, another to crop them, and a third to encode them as biometric data, all embedded in the companion app for a product already at the centre of a mounting privacy crisis.

Meta declined to answer ten questions WIRED posed before publishing, including whether it had already created the database of face profiles NameTag uses, how long the app retains photographs and biometric data of unrecognised people, and whether that data would ever be sent back to Meta’s servers. The company also did not respond to questions about whether it was building NameTag for blind or low-vision users, or to criticism from privacy advocates who warned the system could let stalkers and abusers identify strangers in public.

NameTag first surfaced in February, when The New York Times, citing internal Meta documents, reported that the company was developing face recognition for its smart glasses and considering a launch as early as this year. One internal memo reportedly described releasing the feature during a “dynamic political environment” when privacy and civil liberties advocates would be distracted by other concerns. WIRED subsequently found that much of NameTag’s machinery had been built into the Meta AI app as early as January, months before any public acknowledgement, adding another layer to the company’s pattern of shipping first and disclosing later when it comes to its smart glasses.

Kade Crockford, director of the technology for liberty programme at the American Civil Liberties Union of Massachusetts, said the removal does not undo the original decision to ship the code and pointed to it as evidence that consumer privacy needs stronger legal protection than Congress has been willing to provide. The Massachusetts House of Representatives last week unanimously passed a consumer privacy bill that, if enacted as written, would impose strong enforcement provisions including a private right of action allowing aggrieved users to sue. “State lawmakers need to do their job and step up to protect consumer privacy,” Crockford said.

Meta’s sneaky tactics in slipping the face-recognition code into its smart glasses show exactly why data privacy bills need the teeth of strong enforcement,” Crockford added. “Companies like Meta prioritise their bottom line, so lawmakers need to speak in the only language its C-suite understands.” Whether a code removal prompted by investigative reporting constitutes a victory or merely a tactical retreat depends on what Meta does next, and on whether the regulatory pressure building on both sides of the Atlantic produces enforceable consequences before the feature quietly returns under a different name.



Source link