What a web terminal unlocks — examples edition

The challenge

Normally, switching to a new terminal program is like switching to a new email app: the buttons have moved, your folders and settings didn't come along, and for a week everything feels slightly wrong. That makes trying a new one expensive — so people don't.

The solution

This new app doesn't replace anything. Underneath, it runs the exact same engine your current terminal runs — same settings, same shortcuts, same Claude, all picked up automatically the first time you open it. The only new thing is the window drawn around it. And if you don't like the experiment, you just close the app and open your old terminal. Nothing was moved, changed, or lost — there is nothing to undo.

The challenge

Ted often has four Claude assistants working at the same time, each in its own tab — like four employees working behind four closed office doors. One has finished. One is stuck, politely waiting for Ted to answer a yes-or-no question, and has been waiting for ten minutes. From outside, all four doors look identical. The only way to find out what's going on is to open each door and peek, over and over.

The solution

Imagine a strip down the left side of the window listing all four assistants, each with a status light: green — finished · blue — still working · amber — waiting for you. Under each name, the last thing that assistant said. The one that needs an answer glows. One click and you're standing in front of it. No more going door to door — you glance at the strip the way you'd glance at a row of elevator lights.

The challenge

All day long, ideas get saved into TDPv4 — Ted's app for collecting thoughts, links, and to-dos. But when something gets saved while you're working in the terminal, you can't see it happen. It's like dropping a letter into a mail slot: you assume it landed in the bin on the other side, but you'd have to walk around and check.

The solution

A narrow panel sits along the right edge of the same window, showing today's saved ideas as little cards — each with its picture, just like in the TDPv4 app itself. When a new idea gets saved, whether by you or by Claude, you watch the new card slide in, right there, without leaving what you were doing. The mail slot gets a window in it.

The challenge

Ted has carefully designed his other tools so they all share one look — the same lettering, the same colors, the same feel, like rooms in one well-designed house. The terminal is the exception. It's made by someone else, it looks the way they decided it should look, and the most you can change is the colors and the font — like being allowed to repaint a rented apartment but never move a wall. Every switch between Ted's tools and the terminal feels like stepping out of the house into a bus station.

The solution

Because this terminal is built out of the same materials as a web page, every part of the window is designable — the tabs, the edges, the headers, the panels. It could match Ted's other tools exactly: same lettering, same thin gray lines, same gentle corners. Put a TDPv4 board and a Claude conversation side by side in one window and they'd read as one product, made by one person, on purpose. The terminal stops being the borrowed room and becomes another room in the house.

The challenge

When Claude works on something big, it produces hundreds of lines of text — a running play-by-play of everything it's doing. Somewhere in that mountain, one step failed and printed an error message. By now it's buried hundreds of lines up. Finding it means scrolling and squinting, like looking for one misprinted line in a forty-page receipt.

The solution

Every step that failed gets a small red dot printed beside it in the margin — and the scrollbar on the right shows little red marks where the failures are, like sticky tabs on the edge of a long document. Click a mark, jump straight to that failure. And right next to the error itself, a small button appears: "Ask Claude about this." One click hands that exact error back to Claude and asks it to explain or fix it. No copying, no scrolling, no squinting.

The challenge

Ted's system gives every saved item a three-word nickname, like "repay inventive crackdown," so it's easy to refer to things out loud. But when one of those nicknames shows up in the terminal's text, it's just words on the screen. To actually see the item it refers to, you have to select the words, copy them, and go ask for a lookup — three steps for something you do many times a day.

The solution

The app learns to recognize those three-word nicknames wherever they appear and turns each one into a link — blue and clickable, exactly like a link on a web page. Click "repay inventive crackdown" and the item it names pops open in a panel beside the terminal. Ted's private vocabulary becomes something you can click your way through, the same way you click your way through the web.

The challenge

You give Claude a half-hour job and switch to reading something in another window. Claude finishes early — but nothing taps you on the shoulder. The window just sits there, quietly done, like a dryer in the basement that buzzed once while you were upstairs. Twenty minutes evaporate before you think to check.

The solution

The moment Claude finishes, it sends the app a silent "I'm done" message — silent meaning no text appears anywhere on screen; it travels on that hidden channel. The app turns it into a normal Mac notification, the little banner that slides in at the corner of your screen, and flips that assistant's status light to green in the side strip. It's the dryer learning to text you instead of buzzing once in the basement.

The challenge

Ted has little helper tools that save an idea into TDPv4 straight from the terminal. Handy — but each one prints a "Saved!" message right into the middle of whatever you were working on. Do that a dozen times a day and your workspace fills with clutter, like someone scribbling "received, thanks!" in the margins of the document you're trying to read.

The solution

The helper tool sends its "saved it!" news over the silent channel instead — nothing gets printed into your work at all. Off to the side, the ideas panel flashes a small checkmark and the new card appears. You still get your confirmation, but it shows up in the place built for it, and the text you're actually working in stays clean.

The challenge

Working in a terminal is normally like a phone call: when the window closes, everything said in it is gone. Three weeks ago Ted fixed a tricky problem on his server, step by step, in a terminal window. Today he needs to remember exactly how — and the answer evaporated the moment that window closed. All that's left is memory.

The solution

Every work session saves itself automatically, like a document, when it closes. Type the server's name into a search box and that day's session reopens looking exactly as it did — every instruction Ted gave, every reply the computer made, colors and all. And if a friend asks "how did you fix that?", Ted can send a link to just the three steps that mattered — the way you'd link someone to a paragraph instead of photographing the whole page.

The challenge

You start Claude on a long job and want to go get coffee. The catch: partway through, Claude sometimes stops to ask permission — "is it okay if I change this file?" — and then waits for an answer. If you're out of the room, it just sits there waiting, and the half-hour job becomes a two-hour job.

The solution

Because the whole app is built from the same materials as a web page, the very same screen can open in the browser on your phone. From the coffee line you watch Claude's progress live, and when the permission question appears, you tap "yes" right there. The job keeps moving while you're out. This isn't a separate phone app someone would have to build — it's the same app, reached from a different device, which is a thing web technology simply gets for free.

The challenge

You step away for twenty minutes. While you're gone, Claude carries out thirty separate steps. When you come back, all thirty have run together into one unbroken wall of text — and the only question you actually care about, did anything go wrong?, can only be answered by scrolling from the top and reading everything.

The solution

Because of those invisible bookmarks, the app knows where each of the thirty steps begins and ends — so instead of a wall of text, you come back to a tidy list of thirty steps, each shrunk to a single line with a green or red mark and how long it took. At the top, a small banner: "28 ok · 2 failed." Click it, and the list shows only the two problems, opened up for reading. A wall of text becomes a checklist with a report card stapled to the front.

The challenge

At the end of a busy day, "what did Claude actually do today?" is surprisingly hard to answer. The work happened across several windows, some long closed, and the only records are those walls of text. There's no page anywhere that simply lists the day's work.

The solution

Since the app now knows every step as a tidy fact — what was asked, which project it belonged to, whether it worked, how long it took — it can keep a diary automatically: every step Claude took today, grouped by project, with the problems flagged in red. That diary could even file itself into TDPv4 each evening as a card, so the day's work summarizes itself without anyone writing anything down.