Stackブラウザ: how to decide what you need

Anyone comparing browsers built around multiple web apps runs into the same wall. Every product page describes a calm, organized workspace, and every one of them sounds correct. Feature lists do not separate the candidates because they are written to sound similar. What separates them is duller and far more useful: whether the thing can be obtained, whether it runs on the machines already in play, what the engine underneath decides on your behalf, how the money is structured, what data can be carried in and out, and whether the build is still being worked on. Six checks, in that order, settle most of these decisions in an afternoon.

The order matters. There is no point pricing a product that cannot be downloaded, and no point reading about extensions if the operating system is not supported. Work top to bottom and stop at the first check that rules a candidate out.

Check one: whether it can be obtained at all

The most basic question is whether the software is publicly available. For Stack, the answer is documented on the pricing page.

Stack is currently in a closed beta phase and not yet publicly available. Source: stackbrowser.com

The same entry lists three routes into the beta: users of the earlier product, people who signed up on the waitlist, and buyers of a lifetime license. An invitation code is entered on the start screen to activate the app. Lifetime buyers are told they can get in either through the email address used at purchase or by connecting the wallet tied to the license.

Access that is granted per person rather than per download has practical consequences that are easy to miss during evaluation. Setting up a second machine, handing the same environment to a colleague, or rebuilding a laptop after a failure all depend on that route staying open. A tool that one person can use but nobody else on the team can install is a different proposition from one that anybody can download.

This check also sets expectations about support. Closed beta software is usually shipped with a direct support channel rather than a public help site, which is fine for an individual and harder to justify for a group that needs documented answers. Decide here whether the use case is personal or shared, because everything downstream depends on that answer.

Check two: whether it runs on the machines in play

Supported platforms are the fastest way to eliminate candidates. The pricing FAQ states that Stack runs on Windows and Mac only, with no Linux release, and the answer still carries an estimate of Q4 2023 for a Linux version. That text has not been revised since, so it says nothing reliable about current plans.

The download page offers three builds: macOS on Intel, macOS on Apple Silicon, and Windows on 64-bit. The macOS disk image is roughly 140 MB. On a managed work laptop, the size matters less than the policy: some device management setups block installation of applications that are not on an approved list, and a closed beta build is unlikely to be on one.

Two behaviors are worth testing on day one, because they are the ones that differ even between browsers built on the same engine. The first is a video call with screen sharing, since camera, microphone, and screen recording permissions are granted per application on macOS and do not carry over from another browser. The second is any site that streams paid video, because playback depends on a component that is not present in every build. Both take five minutes to verify and both are expensive to discover in the middle of a meeting.

There is a second half to this check that most comparison articles skip. Count the machines, not the people. A single user with a desktop at the office, a laptop at home, and a personal machine has three installs to keep in step, and three places where settings drift apart. If the tool has no sync, the drift becomes a weekly chore. If sync exists but is tied to a paid tier, the real price of the tool is the paid tier, not the free one.

Check three: what the engine decides on your behalf

Rendering behavior, memory usage, and how quickly security fixes arrive are all decided by the engine, not by the interface. The pricing FAQ is direct about it.

Stack runs on the Chromium engine. Each web card has the same memory and CPU usage as it would in a Chrome tab. Source: stackbrowser.com

The same section explains that the app is built on a forked version of Electron with native components added, and that a card suspension mechanism can put an unused card to sleep to reduce consumption.

Three conclusions follow. Sites that render correctly in Chrome will very likely render correctly here, which removes a whole category of risk for web applications used at work. Memory scales with the number of open cards, so the practical ceiling is the number of things kept open rather than the app itself. And the speed at which the project tracks upstream Chromium releases is what determines how quickly known vulnerabilities get closed, which is worth weighing for a browser that holds work logins.

Extensions belong to this check as well. The same FAQ states that Chrome extensions are not currently supported. For anyone whose password manager, ad blocker, or translation tool is an extension, that single line decides the matter. For anyone using a standalone password manager and system level blocking, it changes nothing. Resolve this before reading any further, because it cuts the candidate list roughly in half either way.

Check four: the shape of the price, before the number

Published pricing for Stack has three shapes rather than one number.

Plan	Price	What the page lists
Starter	Free forever	Unlimited profiles, 5 Flows, 3 color themes
PRO monthly	15 dollars per month	Unlimited Flows, 11 color themes
PRO annual	144 dollars per year	Same scope as monthly
PRO lifetime	199 dollars once	Adds exclusive cursors, an NFT-linked license, and holder perks

Two details carry more weight than the headline figures. Profiles are unlimited on the free tier, which means the ability to stay signed in to several accounts of the same service can be evaluated without paying anything. The paid boundary sits at the number of Flows and the color themes instead.

The lifetime tier is capped at 4,242 licenses, and the counter on the page shows 993 sold. The refund policy states that a request within 14 days is processed within 5 business days, with a separate procedure when an NFT is attached to the purchase: the token has to be listed on a marketplace, and if it has not sold 30 days later the company buys it back at the original price in dollars. A lifetime purchase with conditions attached to the exit is a different financial decision from a monthly subscription, and it deserves to be read in full rather than skimmed.

For comparison, it helps to put the free ceiling of each candidate side by side. A page such as Pricing that lists a free tier alongside monthly, yearly, and one-time options makes it obvious which shape fits a given budget cycle.

Check five: what comes in, and what can leave

Import is the check people run. Export is the check that saves them later.

On import, the FAQ states plainly that migrating data from other browsers is not possible yet, and that history, bookmarks, active sessions, and passwords are being added gradually. Starting from empty is not a disaster, but it is work, and it should be planned rather than discovered on a Monday morning.

Two preparations make any evaluation reversible. Export bookmarks from the current browser to an HTML file and keep it outside the browser. Move passwords out of browser storage and into something independent, so that no single application holds the only copy. With those two done, trying a new tool costs time and nothing else.

There is a third preparation that takes two minutes and pays for itself. Write down every service currently pinned or bookmarked for daily use, with its URL and which account it belongs to. That list is what turns a rebuild from an afternoon of remembering into fifteen minutes of typing, and it is the same list needed to evaluate any other candidate later. Tools that let a service be added by name rather than by URL shorten that step further, and the range covered by such a catalogue, as shown on Supported apps, is a fair proxy for how long the first setup will take.

On export, three questions are worth asking before paying: whether local data remains readable after a subscription lapses, whether there is any export function at all, and whether the list of configured services can be written down somewhere outside the app. A tool that stores everything locally is easy to back up and hard to sync. A tool that syncs through an account is easy to move between machines and dependent on that account staying active. Neither is wrong, but the trade needs to be a decision rather than a surprise.

Check six: whether the build is still moving

The last check is the one nobody can answer from an article, because articles keep their original date while software changes. The artifact itself is the honest source.

Take the download URL from the product page and ask the server about it:

curl -sI "<download URL>" | grep -i last-modified

The last-modified header returns the date the file was placed there. Run against the official Stack builds, both the macOS and the Windows installers return 10 October 2024, and the filename carries the version, 4.9.13 with a beta suffix. That is a fact rather than an opinion, and it can be re-run at any time to see whether it has changed.

Two more probes cost nothing. Open the vendor domain and watch where it lands, since a product that has wound down often redirects before it announces anything. Then check the release notes or changelog page and note the most recent entry. Run all three against every candidate and the shortlist sorts itself into three groups: actively shipping, shipping slowly, and redirected elsewhere.

Putting the six answers on one page

Written out, the six checks fit on a single sheet, and a candidate either has an answer in each row or it does not. An empty row is itself a finding, because it means the vendor has not published something a buyer needs.

Two decisions usually remain once the sheet is filled in. The first is where the boundary between accounts should sit. Separating the view is a visual arrangement; separating the cookie jar is an identity. Anybody who needs two accounts of the same service signed in at once needs the second kind, and the difference is explained concretely on Workspaces. The second is how much of the daily job should move into one window. Work that lives in web apps benefits from consolidation, while work that constantly drops into a shell or a native app can lose time bouncing in and out, unless the tool can reach those too, as described on Built-in terminal.

What to change first

Fill in the six rows for the two strongest candidates and stop reading comparisons. Run each of them as the only browser for one full week, with a real account signed in and real work in front of it, because that is the only test that reveals the habits a feature list cannot describe. If a shortlist is still needed, a product that documents its free ceiling and its separation model openly, as SpaceDeck does, is quicker to evaluate than one that requires an invitation before the first screen.