Thoughts on reading, building software, and the tools I use.
The card that offers a related save from your own queue used to invent 1 connection in every 6. After 5 shipped fixes it invents 1 in 32, it hands you unread saves first, and it stays quiet when nothing truly relates. Finishing one article now shortens your pile instead of just ending.
Switch a noisy newsletter address off in Readplace, then turn it back on whenever you want it — the same address, so every subscription pointing at it keeps landing in your queue. Disabling used to be a one-way door: re-creating the address by name minted a different one and your newsletters broke with nothing to warn you. Every disabled address now carries an Enable button that brings the exact one back to life.
Save a link Readplace already kept and, when its copy had gone stale, the save used to make you wait: it went back to the page, downloaded it again, and re-read it before it answered. It comes back right away now and refreshes the copy in the background, the same on the website, the browser extension, and the iPhone app.
The Readplace browser extension used to make you wait for the whole page to upload before it said Saved. Now it saves your link the instant you click, so you can close the tab a second later and still keep it. The page finishes uploading on its own, in both Chrome and Firefox.
A full page of the Readplace queue is 20 cards. Drawing it used to fire 40 separate database reads, each hauling back the whole article row. Two batched, projected reads now do it, and the archive and delete buttons dropped a read of their own, which closed a race that could leave a broken half-row behind.
Deleting a saved article in Readplace used to be one click with no undo. It asks for confirmation now, built on the browser's native HTML popover with no JavaScript, no extra route, and no change to the delete API the apps and extensions already use.
Readplace saves the article links out of your newsletters as they arrive, so most are in your queue before you open the email. The Save button read Save to queue for all of them anyway. It now marks a saved link as saved, from a fact every save publishes rather than a look into your reading queue.
The Readplace iPhone app has left TestFlight and is on the App Store. Save any page or PDF from the share sheet, read your saved copy with its TL;DR in the app, and sign in without leaving it. It runs on a Mac too. There is still no Android app and no offline reading.
Share a link to a short article and the person who opens it can read the whole thing with no account. Readplace now checks the estimated reading time before its public reader page ever shows a sign-in wall, so a piece you can finish in one sitting stays open for good.
I ran my own reading system on Gmail filters and DynamoDB for ten years. When Pocket and Omnivore shut down, I turned it into a product. Here's the story, and the promise I wrote into the code.
Readplace saves the article links out of your newsletters, and now it clears the utm tracking tags off them too. Only the utm tags go, and only after the page is crawled, because the other parameters can be the part that makes the link open.
Give each newsletter its own Readplace address and the article links inside it land in your reading queue, crawled and summarized. The unsubscribe footers, ads, and menus get dropped, so only the few links worth reading survive, and you can switch off any one address without unsubscribing.
A brand-new reading account used to open on nothing, the article that talked the reader into signing up already gone. Readplace now holds that article on the logged-out reader for two hours and, if you sign up in that window, saves it into the new queue before you add a thing.
Saving a Wikipedia article used to flatten it into one column, with a stray '[edit]' where every section heading had been. The cleaner that strips page clutter was counting the edit link beside each heading as clutter and taking the heading with it. Readplace now clears those edit links before the cleaner scores the page, so the headings survive.
Readplace used to cap a saved file at a few megabytes and send anything larger to a slow path by email. It now takes PDFs up to 500 MB by handing the browser a one-time upload address straight to storage, then checking the staged file before it counts as saved.
Some pages can't be turned into an article, so their AI TL;DR has nothing to write. Readplace's summary pipeline used to retry that job on a loop, about 170 dead-letter messages a month. It now reads the crawl's verdict first and marks the summary done when no text is coming.
Readplace shows the exact date and amount of your next renewal on the account page, but only in the 30 days before it lands. The number is read live from the billing provider when its stored copy goes stale, hidden inside the iPhone app to satisfy App Store rules, and paired with a live cancelling state that heals into a one-tap reactivate.
Readplace re-fetches a page you kept, and when the text comes back changed it stores that version too, dated. The reader bookmark now lists those dates newest first, with the copy you're reading marked best, and every version is held in storage that keeps them.
Connect an app to Readplace on a phone and the fastest login binds the approval to whichever account your browser already holds, with no email shown. The consent screen now names that account and gives you a one-tap way to switch to the one you meant.
Deleting a Readplace account removes the account, every saved article, and the reading history behind it, with no undo. That button used to take one tap on OK. It now takes a phrase you type by hand, checked on the server, so the door out stays yours to open on purpose.
A shortener, a Google News URL, or a newsletter tracker points at the real page through a chain of hops. Walking that chain is part of the save, and Readplace now walks it the same way on every fetch path it uses for sites that block a plain request.
A logged-out reader who tapped Save to My Queue on a Readplace article used to land on the sign-in page, built for people who already signed up. It routes to signup now, so the newest reader meets a form that has somewhere for them to go, and the saved link still rides through to the queue.
Readplace's import page worked, and a noindex tag kept search engines and AI assistants from ever seeing it. It's indexable now, it answers the questions people ask before they start, and it opens on a paste box.
Readplace now shows when each saved copy was captured. A small bookmark rides the right edge of the reader with the date and time it fetched your copy, written in UTC so it reads without JavaScript and rewritten into your own time zone. The live page can change with no note, but the copy holds still at a moment you can now see.
Readplace takes a Sign in with Apple login now, Hide My Email and all. Apple hands the site a forwarding address at privaterelay.appleid.com and keeps your real inbox private. That mask only works if Readplace's mail clears Apple's relay check, so the apex now carries the SPF and DKIM records a human reply needs to reach you.
Readplace used to email you about the first saved article that turned readable and drop the rest for six hours. Now a single digest, sent at most every 6 hours, lists every article you were waiting on that is ready, each with its title, its site, and the first lines of the clean reader text.
Sharing a PDF from an iPhone used to skip Readplace, because the app was missing from the share sheet for any PDF. Fixing how the share extension declares what it accepts brings it back for Safari's viewer, Files, and mail attachments, and it uploads the file's bytes directly so a PDF behind a login still saves.
A saved link from a normally reliable site came back empty while the same page opened in a browser. Readplace's crawler impersonated Chrome 116, whose TLS fingerprint had aged off the list of current browsers that anti-bot edges allow, so it read as a bot. Switching the persona to Chrome 131 turned those sites from 403 back to 200.
The Readplace iPhone app now opens a saved article inside the app, the clean reader and the AI summary, instead of sending you to the source site. Swipe marks an article read and keeps it. Getting the in-app reader to open as you took a server endpoint that trades the app's sign-in token for the session cookie the web reader trusts.
Readplace's link importer is now reachable logged out. Upload a file or paste a link, review the URLs it finds, and pick what to keep before you sign up. The account waits until you commit, and your reviewed selections survive it.
Connect an AI assistant to Readplace and it can read your saved articles: the cleaned reader text and the AI summary, looked up by id. It can also mark one read or unread, the same write the app makes and just as easy to undo. Deleting is the one thing that stays with you.
Connecting your AI assistant to Readplace needs no API key. The MCP client registers itself, you approve one OAuth sign-in, and you can revoke that access anytime. Then your assistant can save pages to your queue and list them back.
Readplace turns away disposable inboxes at signup. A read-it-later list only pays off in the gap between saving and reading, and a throwaway address is gone before the gap opens, so the form asks for an inbox you keep.
Readplace publishes agent-discovery records in DNS, so an AI assistant can find your reading queue's tools the same way a browser finds the site. The zone is signed with DNSSEC, so the agent trusts the answer it gets back.
Save a link to Readplace and it builds a clean copy on its own servers. The source page can change, hide behind a paywall, or go offline, and your saved article stays readable, with no expiry on your copy.
Readplace runs an MCP server. Connect Claude, ChatGPT, Perplexity, or any MCP client, then ask it to save a page to your reading queue or list what you have saved. You log in once with OAuth, and your assistant does the rest.
Readplace now saves LinkedIn posts with their paragraph breaks intact. The page hides its paragraphs inside an inline tag split by double line breaks, and most readers flatten that into one block. Readplace rebuilds the paragraphs, so a saved post reads the way it was written. The same fix helps Substack notes and plain blog posts.
Readplace fetches every link you save on its own servers. Two guards stop a hostile page from reaching our network or slipping code into your reader, and your login now rides https only.
Save an article in Readplace, tap share, and send the link. The preview shows Readplace's clean reader copy with the saved thumbnail, and a tap lands your friend in the quiet version. Search engines still credit the original publisher.
Save a heavy page in Readplace, then close the tab. When the clean reader view and summary finish, Readplace emails you a link. It only emails if you looked and left, and at most once every six hours.
Readplace now publishes standard discovery files so an AI agent can find its API, read docs written for machines, and log in with a safe OAuth flow. A content signal lets AI read the site to answer a question but keeps it out of model training.
The Readplace iPhone app saves from the iOS share sheet. Open any page in Safari, Chrome, or another browser, tap Share, and pick Readplace. The link saves to your reading queue and renders in the background, with no copy-paste and no opening the app first.
Readplace's browser extension used to play a fake loading animation. Now the progress bar tracks the real save: it reads the page in your tab, then uploads it to Readplace, then fills to done. You see each step and know the save worked.
Readplace's reader used to leave a dead video box mid-article. Now each embedded video becomes a short link to the source, so you read the text clean and tap through to the video only if you want it.
Some PDFs sit behind bot protection or a login, so crawlers get turned away. Readplace now reads the PDF from your own browser tab and saves a clean copy to your queue. Open it in your browser, and you can save it.
Paste any link at readplace.com/view and the article opens in a clean reader with a short summary, no signup. Share the link, and anyone reads it the same way. One button saves it to your own queue.
Readplace runs no Google Analytics, no ad pixels, and no tracking scripts. It measures traffic with a salted IP hash and two anonymous first-party cookies that stay on its own servers.
A .txt link, like a Project Gutenberg book or an old internet spec, used to show a 'not a webpage' error even after Readplace had read and summarised it. Now plain text opens in the same reader as web articles and PDFs.
Your Readplace queue now shows a clear Read or Unread label on every article. Mark an item read or unread with one tap, and undo a mis-tap with a single click.
Most read-it-later apps store the text and link the photos back to the source site, so they break when the site changes. Readplace saves a copy of every image on its own CDN. A recent fix makes sure each photo in a multi-image article gets saved, at every screen size.
Readplace keeps one TL;DR per link. When it later reads a cleaner copy of a page, it rebuilds the summary instead of serving the stale one, so your saved articles stay accurate without a manual refresh.
Readplace can now import links from a single page. Paste a newsletter issue or a links page, and Readplace lists every outbound link for you to pick from. Save the whole batch in one click.
Some apps fail to save a link. They fetch the page twice and trip its bot protection. Readplace now fetches each page once, then routes HTML, PDFs, and tweets from that single download. Faster saves, fewer blocks.
A scanned 1950s typewritten PDF broke Readplace's reader view. The fix took two architectures: an LLM-first pipeline that didn't work, a Tesseract pipeline that did, and the LLM-as-post-processor stages that ship today.
Most AI reading tools replace reading with summaries and auto-curation. Readplace uses AI to help you read more of the right things.
The HTTP request already contains the fields that matter. The rest is a hash function and a log line, not a third-party tracker and a consent banner.
Honest comparison of read-it-later apps in 2026: Readplace, Readwise Reader, Instapaper, Raindrop.io, Karakeep, Wallabag, and Matter.
An honest look at what free really means for read-it-later apps, from hosted tiers to self-hosted options — and where a paid subscription with a 14-day free trial fits.
Readplace uses AI to generate short article summaries for triage, not replacement. Here's how it works technically, and why it avoids the usual AI content problems.
Omnivore shut down with two weeks notice. Readplace is a privacy-first read-it-later app built by a developer and no VC funding.
Pocket closed on July 8, 2025. If you missed the export, your articles are often still recoverable. Here's how to find them and move to Readplace.
A fair comparison of two developer-focused read-it-later tools, one self-hosted and one managed, and the tradeoffs each makes.
You open 15 tabs from the HN front page and read 3. The rest haunt your browser for weeks. There's a better system.
You copy articles into ChatGPT for summaries. It works, but it's slow, manual, and nothing gets saved. Readplace does it in one click.
A fair comparison of two read-it-later apps that took different paths after Pocket shut down. One is familiar and stable. The other bets on AI and active development.
Most developers subscribe to far more newsletters than they read. The problem isn't the newsletters. It's the lack of a system to extract the few links worth reading from each one.
A fair look at two read-it-later apps with AI. Where Readwise wins, where Readplace wins, and how to pick the right one for you.