Build a private family schedule app that emails caregivers with Claude Code

Find forgotten online accounts from old signup emails with Claude Cowork and other AI agents

Ben Patterson wanted to know how many forgotten accounts were left over from years of testing products. For PCWorld, Patterson gave a prompt to Claude Cowork, Gemini Spark, Meta's Muse and a ChatGPT Dot: search the last 5 years of email for welcome, verification and password-reset messages, build a spreadsheet of services with first and last email dates, whether the service is on haveibeenpwned.com's list of breached sites, and its official account-deletion link, with the most dormant at the top, and don't sign in, reset passwords or contact anyone. All four returned spreadsheets in five minutes to half an hour; each found at least 50 accounts, and the ChatGPT Dot found 73. Patterson found accounts they'd forgotten, such as a Bowers & Wilkins account from 2022. About half the deletion links came back 'not found' or pointed to privacy FAQs.

Grab an appointment on a fully booked site by having ChatGPT Work check every 30 minutes

Yuvraj Kashyap set ChatGPT Work to check a busy appointment booking site every 30 mins. After 6 days, it found an opening and booked it. Kashyap called it the first practical use that genuinely saved a lot of time and attention.

Work out why a meeting was missed and set up a tested email-to-calendar check with ChatGPT and Codex

Dave W Plummer missed an important appearance and asked the ChatGPT desktop app, which has access to their Gmail, to work out how it happened. It found Plummer had noon in mind, the time of an earlier prep call, while the event started at 7:50 a.m. Pacific, with emails also mentioning 8:50 a.m. Mountain. When Codex then reconstructed the email trail, it found the organiser had cancelled the calendar invite sent to Plummer's regular address while sorting out a Teams login mismatch, and the later confirmation email had no new invite. With Codex, Plummer then set up a recurring agent that reads email threads, including changes and cancellations, adds missing calendar entries with the source email and link, and flags contradictions. Test alerts at first didn't notify on the iPhone; after checking alarms, sync and Focus settings, a test finally rang. It now checks every ten minutes while the Mac is awake; its first historical review was still running.

Sort 204 unread emails into action labels with ChatGPT, approving every change first

Ary Aranguiz connected ChatGPT to Gmail and asked it to create four labels: ACTION/Reply, ACTION/Decision, ACTION/Review and STATUS/Waiting. ChatGPT then reviewed 204 unread emails from the past seven days and proposed which belonged under each label, separating people waiting for a reply, invitations needing a decision, information to review and routine mail needing no action. It noticed the original label set was incomplete and recommended adding ACTION/Review. Aranguiz approved the proposals, and only then did it create the label and apply the labels. It didn't send, delete, archive, forward or mark anything as read.

Catch a returned $2,350 insurance reimbursement check buried in 80k unread emails with Grok Bot

Within five minutes of setting up Grok Bot, a poster with about 80k unread emails had it flag one message from four days earlier: the insurance company said their windshield replacement reimbursement check had been returned by the post office and needed a new address. The poster says it was the only email the bot flagged and that it saved them $2350.

Fill three web onboarding forms from Google Drive files with Claude in Chrome, at a heavy usage cost (partial)

Evan Solomon asked Claude in Chrome to fill out three web onboarding forms using information scattered across their Google Drive. It worked very well: it read the pages, found the information in Drive, filled the fields and handled multi-step forms with almost no hand-holding. But it took 574 model round-trips and about 223M tokens, roughly $350–400 in API-equivalent compute, and used up the Max plan's 5-hour window in one session. Solomon notes it was slow, got slower and pricier as the session grew, and that running it on the top model was a mistake for form-filling.

Build a points-for-chores reward app for four kids with Claude Code

A Japanese dad who blogs as ht and is raising four boys spent about a month chatting with Claude Code to build a chore web app, using GitHub, Node and Supabase for the first time. Parents list the chores they want done and the rewards; kids tap a 'Done! Check' button to earn points, such as 100 points for washing dishes or hanging laundry, and trade them for rewards like a 100-yen snack. Fights over chores led to a view showing only each child's chores, and a one-tap on/off switch made daily setup easy. The kids began choosing chores each morning and even helping with chores that weren't listed. The favorite reward was a night of Dad sleeping beside them. ht said the family uses it less than at first, but the kids are still saving points.

Build a summer activity checklist app for a child with Claude Code

Suri, a mom who blogs at Dreaming Loud, wrote a list of summer activities an older child, V, would enjoy and used Claude Code inside VS Code to turn it into an app with a calendar and progress bars. Activities are grouped by month for June, July and August and range from a full day outdoors to learning a headstand and 15 minutes of reading and math worksheets most evenings. A calendar button adds any scheduled activity to Google Calendar, the activity images come from Pexels, and V checks items off on a phone or laptop as the bars for the summer and each month fill up. Suri said seeing V's smile when the bar moves is the motivation to follow through.

Decode a scary dental insurance denial and confirm the real bill with Claude Code

John Sambrook, 66, woke at 4:30 a.m. to an Aetna Explanation of Benefits for some dental fillings: six of the eight line items said DENIED, and the 'Your Share' column added up to $1,316. Sambrook had spent about 15 minutes setting up a small Claude Code workspace for personal insurance, with the policy documents in it, and asked it to explain the EOB. About three minutes later, it showed that the plan excludes 'comprehensive dental services', so the denials were expected, but that the dentist is in network and bills at a contracted rate. The actual amount owed was $723.83, the figure already being paid, so nothing more was due.