@admin — following
Follow buttons here act as you, not as @admin.
@admin follows
| Label | Kind | State | Action |
|---|---|---|---|
| indiekit-demo.rmendes.net | source | — | — |
| fediversereport.com | source | — | — |
| wedistribute.org | source | — | — |
| hnrss.org | source | — | — |
| hnrss.org | source | — | — |
| hnrss.org | source | — | — |
| hnrss.org | source | — | — |
| hnrss.org | source | — | — |
| hnrss.org | source | — | — |
| hnrss.org | source | — | — |
| hnrss.org | source | — | — |
| 9to5google.com | source | — | — |
| hnrss.org | source | — | — |
| hnrss.org | source | — | — |
| cvefeed.io | source | — | — |
| thebeatcommunity.com | source | — | — |
| contabo-status.com | source | — | — |
| www.techdirt.com | source | — | — |
| beehaw.org | source | — | — |
| beehaw.org | source | — | — |
Timeline
-
🟢 PR #1001 — fix(frontend): load the app script once the page is parsed
Reply sourcePull request (OPEN) · PR #1001 fix(frontend): load the app script once the page is parsed · @rmdes
Fixes #1000
Drops
asyncfrom the app script so it runs after the document is parsed, and custom elements are defined once their content exists. A module script is deferred by default, sodeferis kept only to say so.Checked with the proxy reproduction from the issue: five errors and no add-another buttons with
async defer, none and both buttons withdefer. -
🐛 Issue #1000 — Frontend components can start before their content is parsed
Reply sourceIssue ouverte · Issue #1000 Frontend components can start before their content is parsed · @rmdes
The app script is loaded with
async defer(packages/frontend/layouts/default.njk:46). For a module scriptasyncwins, so the script can run while the page is still being parsed. It defines the custom elements, and an element that has been opened but not yet filled getsconnectedCallbackstraight away:add-anotherlooks for#add-button(components/add-another/index.js:27) before it exists,getElementthrows, and the component never sets up.It depends on timing, so it shows up on some loads and not others. To make it repeatable, I sent
/posts/create?type=photothrough a proxy that pauses for 1.5 s after the<add-another>start tag. Withasync defer:No element matching #add-button No element matching .file-input__progress No element matching label No element matching .label No element matching .geo-inputand no add-another buttons. With
deferalone: no errors, both buttons.asynccame in with Stimulus (b9d05da), which waited for the document before connecting controllers; it became a problem when the components moved to custom elements (e.g. 914c8fd). -
📝 24b659a — fix(frontend): don't drop attributes after one without a value
Reply sourceCommit sur main · 24b659a · @rmdes
attributes() returned on the first falsy value instead of skipping it, so everything after it was lost. The summary field’s editor-locale is empty when application.locale is unset, which dropped editor-status=“false” and editor-toolbar=“false” after it and turned the summary into a full editor.
-
Genius or Mid Life Crisis Lol
Reply sourceArticle URL: https://www.reddit.com/r/founder/comments/1x25aws/genius_or_midlife_crisis/
Comments URL: https://news.ycombinator.com/item?id=50038133
Points: 1
# Comments: 0
-
🟢 PR #999 — fix(frontend): don't drop attributes after one without a value
Reply sourcePull request (OPEN) · PR #999 fix(frontend): don’t drop attributes after one without a value · @rmdes
Fixes #998
Skips an attribute without a value instead of returning, as 37e15d5 intended. Adds a test with an empty attribute followed by others (fails before the change, passes after).
-
🐛 Issue #998 — Frontend attributes() drops every attribute after one without a value
Reply sourceIssue ouverte · Issue #998 Frontend attributes() drops every attribute after one without a value · @rmdes
attributes()inpackages/frontend/lib/globals/attributes.jsreturns as soon as it meets a value that is empty or false, so every attribute after it is lost. The earlyreturncame in with 37e15d5 (January 2023, “don’t return attributes with falsy values”);continuewas meant.The article summary shows it.
summary-field.njkpasses"editor-locale": application.localebefore"editor-status": "false"and"editor-toolbar": "false".application.localehas no default (onlypublication.localedoes), so on an install that doesn’t set it, the twofalseattributes are dropped and the one-line summary gets the full editor toolbar and status bar. -
🟢 PR #997 — feat(frontend): choose an existing media file from the editor and file fields
Reply sourcePull request (OPEN) · PR #997 feat(frontend): choose an existing media file from the editor and file fields · @rmdes
Fixes #996
Adds a “Browse media…” button to the editor toolbar and to the file input. It opens a
<dialog>listing files from the media endpoint (q=source, with “Show more” followingpaging.after). Choosing a file insertsfor a photo or[filename](url)otherwise in the editor, or fills the file input’s path.- The dialog markup and strings come from a
<template>rendered by the component, so they are localised; the newmediaBrowserkeys are in every frontend locale. - The editor button shows wherever the editor has a toolbar and a media endpoint, not only where image upload is enabled, so notes get it too. Plain text fields like an article’s summary are unchanged.
- One new toolbar icon (
browse-media), alongside the existing ones. - Tests:
test/unit/utils/media-browser.jscovers listing, choosing, paging, the empty and error messages, and cancelling.
Tested on the quickstart stack with an image built from this branch, in English and French: the photo field, and the article, note and photo content editors.
- The dialog markup and strings come from a
-
Build your own decision model
Reply sourceArticle URL: https://nishtahir.com/build-your-own-decision-model/
Comments URL: https://news.ycombinator.com/item?id=50037949
Points: 451
# Comments: 102
-
📝 88287f1 — fix(frontend): don't throw looking up elements that may be absent
Reply sourceCommit sur main · 88287f1 · @rmdes
691b359 made component lookups throw when an element is missing. Four lookups are of elements that are absent by design, each already followed by a null check: the file-input picker and the geo-input button live only in a until connectedCallback clones them in, a tag-input hint and error message are only rendered when given, and the textarea status bar can be turned off. The throw aborted connectedCallback, so the upload button on photo, video, audio and featured fields and the find location button were never wired, and the category field rendered as a plain input. Those four go back to querySelector with nullable types. New component tests mount file-input and geo-input as custom elements in jsdom.
-
📝 5637cf4 — feat(indiekit): show more configured endpoints in homepage guidance
Reply sourceCommit sur main · 5637cf4 · @paulrobertlloyd
feat(indiekit): show more configured endpoints in homepage guidance
-
📝 07f959e — feat(endpoint-auth): use url filter for endpoint urls
Reply sourceCommit sur main · 07f959e · @paulrobertlloyd
feat(endpoint-auth): use url filter for endpoint urls
-
📝 526ed5d — feat(indiekit): make homepage guidance simpler and remove mention of …
Reply sourceCommit sur main · 526ed5d · @paulrobertlloyd
…Micropub By removing mention of Micropub, we can then show more endpoints that can be included in a website’s head
-
📝 0f8640e — feat(frontend): use url filter for absolute urls in default layout
Reply sourceCommit sur main · 0f8640e · @paulrobertlloyd
feat(frontend): use url filter for absolute urls in default layout
-
🐛 Issue #996 — Choose an existing media file from the editor and file fields
Reply sourceIssue ouverte · Issue #996 Choose an existing media file from the editor and file fields · @rmdes
Today the only way to reuse an uploaded file is to copy its URL from
/files.I’d like to add a “Browse media” button next to the editor’s image upload button and next to file-input’s “Upload file” button. It would open a dialog listing the media endpoint’s files (
q=source, paged by uid) and insert or fill in the chosen one. Built the way the other frontend components are: a native<dialog>, localised strings in the component template, the existing icons.I’ve run a version of this on my own site for a while; a screenshot of it follows.
-
🟢 PR #995 — fix(frontend): don't throw looking up elements that may be absent
Reply sourcePull request (OPEN) · PR #995 fix(frontend): don’t throw looking up elements that may be absent · @rmdes
Fixes #994.
Those four lookups go back to
querySelector, with types that admitnull; every othergetElementcall is unchanged. New tests mount file-input and geo-input as custom elements in jsdom and fail without the fix. They are the first component tests in the package, so tell me if you’d rather they live elsewhere.Found while running a build of
main; worth having before beta 30. -
🐛 Issue #994 — Frontend components drop the upload, find location and tag input set-up on main
Issue ouverte · Issue #994 Frontend components drop the upload, find location and tag input set-up on main · @rmdes
Since 691b359, component lookups throw when an element is missing. Four lookups are of elements that are absent by design, each followed by a null check the throw now makes unreachable:
file-input/index.js:48andgeo-input/index.js:30: the picker and the button live only in a<template>untilconnectedCallbackclones them in.tag-input/index.js:7-8: a hint and an error message are only rendered when given.textarea/index.js:171: the status bar can be turned off witheditor-status="false".
The throw aborts
connectedCallback, so onmainthe upload button next to photo, video, audio and featured URL fields, the find location button, and the category tag input are never set up. Nothing is lost (the URL fields and the editor’s own image upload still work), but those three are missing.Reproduction: open
/posts/create?type=photo. There is no “Upload file” button next to the photo URL field, and the console showsNo element matching .file-input__picker,No element matching .geo-input__buttonandNo element matching .error-message.
Reply source -
Secret of Orange Cats Revealed After 60-Year Search
Reply sourceArticle URL: https://www.sciencealert.com/secret-of-orange-cats-finally-revealed-after-60-year-search
Comments URL: https://news.ycombinator.com/item?id=50037711
Points: 101
# Comments: 75
-
GPT-Astra NFL Player Video Analysis
Reply sourceArticle URL: https://twitter.com/robsandhu/status/2108987321786007853
Comments URL: https://news.ycombinator.com/item?id=50037218
Points: 2
# Comments: 0
-
🟢 PR #993 — Use `canonicalUrl` filter in Nunjucks templates, and show all configured endpoints on homepage
Reply sourcePull request (OPEN) · PR #993 Use
canonicalUrlfilter in Nunjucks templates, and show all configured endpoints on homepage · @paulrobertlloydSomewhat related improvements:
- Use
canonicalUrlfilter in all Nunjucks templates where an endpoint’s mount path (or frontend asset path) needs to be prefixed with the application URL - Remove the
getEndpointUrlsmethod that was used to provide a list of endpoint mount paths for use on the homepage guidance - Update home page guidance to list all
s that can be added to a website’s` to support installed/configured endpoints - Improve the wording of that guidance to be a bit less technical, and remove specific mention of Micropub now that we can show links for other endpoints
- Use
-
A city-building game in which the city would prefer you didn't
Reply sourceArticle URL: https://housing.over.pizza/
Comments URL: https://news.ycombinator.com/item?id=50036864
Points: 530
# Comments: 189
-
There has been no increase in car sales in China in over a decade
Reply sourceArticle URL: https://twitter.com/lymanstoneky/status/2107227592151445963
Comments URL: https://news.ycombinator.com/item?id=50036400
Points: 5
# Comments: 1
-
Proof-of-Human
Reply sourceArticle URL: https://twitter.com/bzamayo/status/2108498107184934914
Comments URL: https://news.ycombinator.com/item?id=50036394
Points: 3
# Comments: 0
-
What if it could learn and update its own weights as it's being used?
Reply sourceArticle URL: https://twitter.com/ashxhart/status/2108580126094963038
Comments URL: https://news.ycombinator.com/item?id=50036277
Points: 1
# Comments: 1
-
Where Google Wallet state IDs are available on Android
Reply source
States across the US are slowly rolling out support for adding your driver’s license or ID to the Google Wallet app on Android.
more… -
This Week In Techdirt History: October 4th – 10th
This Week in 2016 This Week in 2011 This Week in 2006Reply source -
Nvidia in talks to acquire US 'open' model startup Reflection AI
Reply sourceArticle URL: https://www.ft.com/content/052610c5-22b4-4dd4-932e-b7f9f0628b6a
Comments URL: https://news.ycombinator.com/item?id=50035886
Points: 170
# Comments: 108
-
📝 721f4dd — chore: update dependencies
Reply source -
Vibe coded browser ports of Halo, The Simpsons: Hit And Run, GTA work well
Reply source -
📝 f8c7376 — feat(endpoint-syndicate): syndicate every post awaiting syndication
Reply sourceCommit sur main · f8c7376 · @rmdes
A request without source_url syndicated one post, the oldest awaiting syndication, and stopped. The endpoint exists to be pinged after a deploy or on a schedule, so a backlog of several posts, which is what a deploy that publishes more than one post or an outage at a target leaves behind, needed as many pings as posts. Work through every post that still has a target to reach, oldest first and one at a time, recording each result on its post as before, and respond with a result per post. A target that fails is noted on its post and the batch carries on; an error updating a post stops the batch, and the posts not reached are still awaiting syndication for the next request. A request arriving while a batch is running is told so rather than syndicating the same posts again, since a deploy webhook and a scheduled request can coincide. Two tests sent the post URL as syndication.url, which the endpoint has never read (the form sends syndication.source_url, as the README documents). They passed because the fallback happened to pick the only post; they now send the documented parameter. The mock content store accepts more than one post so a test can seed a backlog.
-
Why DuckDB 2.0 is faster
Reply sourceArticle URL: https://motherduck.com/blog/why-duckdb-20-is-faster/
Comments URL: https://news.ycombinator.com/item?id=50035530
Points: 196
# Comments: 61
-
📝 54f193e — feat(endpoint-share): let the share form choose syndication targets
Reply sourceCommit sur main · 54f193e · @rmdes
The share form sends a bookmark to the Micropub endpoint but offers no way to say where it should be syndicated, so a shared bookmark never reaches a syndicator unless the target is edited afterwards. Render a checkbox per configured target, as the post form does, and forward each chosen one as its own mp-syndicate-to parameter: URLSearchParams would otherwise join an array into one comma-separated value the Micropub endpoint does not match. A target whose info cannot be read is shown disabled with the reason, the same as the post form since the syndicator-info change. Tests load the two syndicators the test configuration already knows about and check the form renders a checkbox per target, that both chosen targets come back as separate values on the created post, and that a chosen target stays checked when the form is re-rendered with errors.
-
FreeBSD Cope with Signal Fuckup
Reply sourceArticle URL: https://vermaden.wordpress.com/2026/10/10/freebsd-cope-with-signal-fuckup/
Comments URL: https://news.ycombinator.com/item?id=50035448
Points: 3
# Comments: 0
-
📝 21624e4 — feat(endpoint-share): let the share form choose syndication targets
Reply sourceCommit sur main · 21624e4 · @rmdes
The share form sends a bookmark to the Micropub endpoint but offers no way to say where it should be syndicated, so a shared bookmark never reaches a syndicator unless the target is edited afterwards. Render a checkbox per configured target, as the post form does, and forward each chosen one as its own mp-syndicate-to parameter: URLSearchParams would otherwise join an array into one comma-separated value the Micropub endpoint does not match. A target whose info cannot be read is shown disabled with the reason, the same as the post form since the syndicator-info change. Tests load the two syndicators the test configuration already knows about and check the form renders a checkbox per target, that both chosen targets come back as separate values on the created post, and that a chosen target stays checked when the form is re-rendered with errors.
-
📝 2f89cd1 — feat(endpoint-posts): pre-fill the create form from url and name quer…
Reply sourceCommit sur main · 2f89cd1 · @rmdes
…y parameters A bookmarklet, a share sheet or a reader’s ‘Post’ button has a URL and usually a title to hand over, and the share endpoint already accepts them. The create form did not, so a client had to choose between the share form’s fixed bookmark and an empty post form. Accept url and name on /posts/create. The URL goes to the field the post type has for one (bookmark-of, in-reply-to, like-of or repost-of) and to the content where the type has none; the name only where the type has a name field. Anything already in the form wins, so a re-rendered form is not overwritten.
-
📝 b792c1a — feat(endpoint-micropub): filter the list of posts by text and category
Reply sourceCommit sur main · b792c1a · @rmdes
The README promises
filterfor every list query and gives/micropub?q=source&filter=webas the example, but the source query ignored it: the parameter only ever reached the configuration queries. Honour it for posts as a case-insensitive match on the name and content, and acceptcategoryalongside it for posts carrying that category. Filtering keeps paging: getCursor takes an optional query that is merged into the range condition and into the hasNext/hasPrev probes, so a filtered page reports whether more filtered items exist rather than whether more items exist at all. Existing callers are unaffected; the parameter defaults to an empty object. Tests cover the cursor filter in util, and for the endpoint: match by name, match by content, match by category, both together across two pages, and no match. -
📝 77a6266 — fix(endpoint-micropub): normalise categories on write
Reply sourceCommit sur main · 77a6266 · @rmdes
A category arrives as typed. Surrounding whitespace, an empty value from a blank form field, and the same word in another case each became a distinct category on the site, and every list built from categories, the q=category query included, grew a near-duplicate for it. Trim each value, drop the empties, and fold case-insensitive duplicates onto the first spelling given, at the point where properties are normalised so every write path is covered. A category left with nothing is removed rather than stored as an empty array. Casing is otherwise left as the author wrote it: folding onto a canonical spelling is a separate question.
-
Models as Insider Risks in the Super Intelligence Era
Reply sourceArticle URL: https://twitter.com/satyanadella/status/2108931348857827686
Comments URL: https://news.ycombinator.com/item?id=50035193
Points: 7
# Comments: 0
-
📝 47b995d — fix(frontend): close the tabindex clause in add-another's focusable s…
Reply sourceCommit sur main · 47b995d · @rmdes
…elector The last clause of the selector was missing its closing parenthesis. Browsers recover from the error, which is why nothing visibly broke, but a selector that only works through parser error recovery is one engine update away from throwing, and it misleads anyone reading it.
-
💬 @paulrobertlloyd — PR #992 : feat(indiekit): declare a collection's indexes with addCollection
Reply sourceCommentaire de pull request · PR #992 feat(indiekit): declare a collection’s indexes with addCollection · @paulrobertlloyd
Brill, thanks @rmdes!
-
📝 df0b8f8 — feat(indiekit): declare a collection's indexes with addCollection
Reply sourceCommit sur main · df0b8f8 · @rmdes
A plug-in declares the indexes a collection needs where it adds the collection, as an array of a key plus any createIndex options, and the core creates them once the database is connected and before serving. The literal posts and media list in index.js goes: the Micropub and media endpoints declare the properties.uid index themselves, and the Microsub endpoint declares its three instead of creating them at init.
-
📝 e8bc4df — feat(endpoint-microsub): add localised strings
Reply sourceCommit sur main · e8bc4df · @paulrobertlloyd
feat(endpoint-microsub): add localised strings
-
🟢 PR #992 — feat(indiekit): declare a collection's indexes with addCollection
Reply sourcePull request (OPEN) · PR #992 feat(indiekit): declare a collection’s indexes with addCollection · @rmdes
Fixes #979.
Indiekit.addCollection(name, indexes)takes an optional array of indexes, each akeyplus anycreateIndexoptions such asunique. The core creates them increateIndexes(), called fromserver()after the plug-ins are installed and before serving; creating an existing index is a no-op, so it runs on every start.The literal posts and media loop in
index.jsgoes: the Micropub and media endpoints declare{ key: { "properties.uid": 1 } }where they add their collections, and the Microsub endpoint declares its three (channel + id unique, channel + uid unique, channel + url) instead of creating them at init. The uid backfill stays where it was. Docs page updated.Two core tests (declared indexes are created, a collection declares none by default); the Microsub index test moves with the declaration.
-
🟢 PR #991 — fix(endpoint-micropub): honour `filter` on `q=source`, add `category`
Reply sourcePull request (OPEN) · PR #991 fix(endpoint-micropub): honour
filteronq=source, addcategory· @rmdesFixes #990.
filteris a case-insensitive match on a post’s name or content,categoryan exact category; both are cast to strings so nothing reaches the database as an operator, and both are passed togetCursoras its filter so paging keeps working within the match (the cursor’sfilteroption landed with #944).Four integration tests, seen failing first. README documents
category.This is one of the last six changes ported from my fork; nothing here needs to hold a release.
-
🐛 Issue #990 — `q=source` ignores the documented `filter` parameter
Reply sourceIssue ouverte · Issue #990
q=sourceignores the documentedfilterparameter · @rmdespackages/endpoint-micropub/README.md:41says list queries supportfilter,limitandoffset.lib/controllers/query.jsreadsfilter(line 32) but applies it only to configuration queries (line 121); the posts list returned byq=sourcenever sees it.Reproduction:
/micropub?q=source&filter=webreturns every post. -
🟢 PR #989 — feat(endpoint-syndicate): syndicate every post awaiting syndication per request
Reply sourcePull request (OPEN) · PR #989 feat(endpoint-syndicate): syndicate every post awaiting syndication per request · @rmdes
Fixes #988.
A request without
source_urlworks through every published post that still has a target, oldest first, one at a time, and responds with a result per post; a second such request arriving while one runs is told syndication is already in progress. A request withsource_urlbehaves as before.On the way: two existing tests sent
syndication.url, a field the endpoint never read, and passed by coincidence; they now sendsource_url. The mock store replies persist so a batch can write several files.This is one of the last six changes ported from my fork; nothing here needs to hold a release.
-
🐛 Issue #988 — A syndication request without `source_url` handles one post only
Reply sourceIssue ouverte · Issue #988 A syndication request without
source_urlhandles one post only · @rmdesgetPostDatainpackages/endpoint-syndicate/lib/utils.js:30-33takeslimit(1)and returns the oldest post awaiting syndication, so a deploy webhook or a scheduled request after a build syndicates one post per call. With several posts published between builds, the rest wait for as many further requests.Reproduction: publish two posts with
mp-syndicate-to,POST /syndicateonce: the first is syndicated, the second still has its target. -
🟢 PR #987 — feat(endpoint-posts): pre-fill the create form from the query string
Reply sourcePull request (OPEN) · PR #987 feat(endpoint-posts): pre-fill the create form from the query string · @rmdes
Fixes #986.
urlgoes to the post type’s URL field (bookmark-of,in-reply-to,like-oforrepost-of), or to its content when the type has none;nameonly where the type has a name field. Anything already in the form wins. README gains a section mirroring the share endpoint’s.Three integration tests, seen failing first.
This is one of the last six changes ported from my fork; nothing here needs to hold a release.
-
🐛 Issue #986 — Post creation form cannot be pre-filled from the query string
Reply sourceIssue ouverte · Issue #986 Post creation form cannot be pre-filled from the query string · @rmdes
packages/endpoint-posts/lib/middleware/post-data.js:20reads onlytypefrom the query. The share endpoint documentsurl,nameandcontentpre-filling, so a bookmarklet or a reader’s post button can hand over a page; the posts endpoint cannot, which rules out a pre-filled reply, like, repost or bookmark.Reproduction:
/posts/create?type=bookmark&url=https://example.website/article&name=An+articlerenders an empty form. -
🟢 PR #985 — fix(frontend): close the tabindex clause in add-another's focusable selector
Reply sourcePull request (OPEN) · PR #985 fix(frontend): close the tabindex clause in add-another’s focusable selector · @rmdes
Fixes #984.
One character: the missing
).This is one of the last six changes ported from my fork; nothing here needs to hold a release.
-
🐛 Issue #984 — add-another's focusable selector is missing a closing bracket
Reply sourceIssue ouverte · Issue #984 add-another’s focusable selector is missing a closing bracket · @rmdes
packages/frontend/components/add-another/index.js:3ends with[tabindex]:not([tabindex="-1"]and no closing). Browsers recover from the malformed last clause today, which is why nothing visibly breaks, but the selector as written is invalid and thetabindexclause is not reliably applied when focusing a newly added item.