Android Feed

← The 5 stories that mattered

Everything collected ·

All 406 items collected for this issue

New here? How to read this issue
  1. Read the picks. Each tab opens with 4–5 stories ranked by how much you’d regret missing them. That takes about 3 minutes. “Start the read” shows one story at a time, and “Listen” reads them aloud.
  2. Stop at “You’re caught up”, or go deeper. Everything after it is the full list of what was collected today, folded away.
  3. Android comes first. Product & AI covers what builders are launching and discussing.
Stable · Beta · Alpha
Release stages. Stable is safe to ship, beta’s API is frozen but may still have bugs, alpha is experimental.
AndroidX
Google’s Jetpack libraries (Compose, Paging, Room…), released separately from Android itself.
AOSP insider
Code changes merged into Android’s open source repository. They often show an API before it’s announced.
Lanes
Where an Android item comes from: official, AOSP insider, experts, open-source library releases, leaks and ecosystem, community.
Adopt now · Watch · Rumour
What to do about it, and how sure the source is.
Dotted underline
An inferred claim: Claude’s reading, not something the source says.

Collected automatically each morning from 37 sources. Notes are written by Claude. Links show what was trending, not endorsements.

Android5Product & AI4
Start the Android read5 stories, ranked by what you’d regret missing · 2 min read

  1. Next Android release ends direct contact writes for new apps

    If your app adds or edits contacts, raising your target SDK will break that code unless you move to the new approach first.

    Google's Android Developers Blog says apps that target the upcoming Android release will no longer be able to write to contacts directly with the WRITE_CONTACTS permission. The post, titled a more reliable way to add and edit contacts, describes the path apps should use instead. Apps that create, import or sync contacts need to move before they raise their target SDK.

    Android Developers Blog · OfficialRead the source

  2. Android 17 reaches about one percent of devices after four months

    Features built only for Android 17 will reach almost no users this year, so plan fallbacks for older versions.

    Google has updated its Android distribution numbers, 9to5Google reports. Android 17, this year's release, runs on about one percent of devices four months after launch. Android 16 now makes up the biggest share, at over 25 percent. These are the numbers most teams use to decide which Android versions to support.

    9to5Google · Android · EcosystemRead the source

  3. Kotlin 2.5.0 reaches its second beta

    A second beta is the time to try Kotlin 2.5 on your project and report compiler problems before the stable release.

    JetBrains' latest Kotlin roundup says Kotlin 2.5.0 is taking shape with its second beta. The same post names the Shipaton 2026 winners, where three Kotlin Multiplatform apps, which share code across Android, iOS and other platforms, took the Ship Kotlin Everywhere Award. It also shares where KotlinConf 2027 will be held.

    Kotlin blog · OfficialRead the source

  4. Android Auto video apps are still coming soon, Google says

    If you planned a car version of a video or browser app, keep it on the roadmap but do not promise users a date yet.

    At its latest developer conference, Google said in a brief update that video and browser apps for Android Auto, the system that puts phone apps on a car screen, are still coming soon. 9to5Google notes that Google's own deadline for them is close. Developers of streaming or browser apps still have no firm date for car support.

    9to5Google · Android · EcosystemRead the source

  5. How to stop flaky Turbine tests for Kotlin Flows

    Most Android apps expose state as Flows, so reliable Flow tests protect the code users see most.

    Kevin Schildhorn at Touchlab published part two of his series on hardening tests. This part covers Turbine, a small library for testing Kotlin Flows, which are streams of values that arrive over time. Flow tests are confusing to write and often flaky, and the post explains how to avoid that flakiness when you test with Turbine.

    Touchlab (Kotlin Multiplatform) · ExpertRead the source

Also worth knowing

Releases, policy and leaks an Android engineer should know about this week.

  • KSP 2.3.13

    KSP 2.3.13 adds KSContextParameter and KSFunctionDeclaration.contextParameters, so symbol processors can read Kotlin context parameters, and improves diagnostics.

    Teams writing or using KSP processors on code that uses context parameters.

    Adopt nowOfficial

Go deeper: every Android item, by lane

Official & releases

57Act on the contacts change first: apps targeting the next Android release lose direct writes with WRITE_CONTACTS; also AppFunctions reached beta, Camera 1.7 is a release candidate, the XR ARCore, Runtime and SceneCore libraries are at 1.0.0, and Kotlin 2.5.0 is in its second beta.

AndroidX library releases

Stable · safe to ship · 8

Beta · API frozen, try it before it ships · 13

Alpha · experimental, APIs may change · 23
Show 5 more

AOSP / AndroidX insider

23Android XR's scene code now batches changes and drops manual transactions, Remote Compose hides its internals and gains a surface modifier for glasses UI, and a WebView change points to sharing memory buffers between Java and web content.
  • CameraX: default check for extension supportnew

    Add default implementation for CameraMetadata#isExtensionSupported

    A default implementation means fakes and custom metadata classes no longer have to override it, API polish while Camera 1.7 is at release candidate.

  • Android XR: scene changes batched togethernew

    Add a coalescing committer for OpenXR scene graph changes

    A coalescing committer groups many scene graph edits into fewer OpenXR commits, performance work for SceneCore now that it is 1.0.

  • Android XR: manual scene transactions removednew

    Remove OpenXrTransaction and cancelSceneTransaction

    Paired with the batching change, apps no longer open or cancel scene transactions by hand; the runtime manages them.

  • Feathered edges: equality and a zero-size guardnew

    Add feathering effect equality and divide by 0 protection

    The soft-edge feathering effect gets equality checks, which avoid needless updates, and a divide-by-zero guard against crashes at zero size.

  • Glasses UI: surface modifier for remote layoutsnew

    RC Glimmer: Add RemoteModifier.surface API

    Remote Compose meets Glimmer, the design system for Android XR display glasses, a sign that glasses UI may be drawn remotely from a phone app.

  • Compose Remote: draw paths from raw arraysnew

    Add addPathData(float[], int) overload to RemoteComposeWriter

    A lower-level path writer keeps vector shapes compact in remote UI documents.

  • Dialogs: tunable gesture thresholdsnew

    Add gesture thresholds to AlertDialogDefaults

    A large change adding gesture thresholds to AlertDialogDefaults suggests dialogs that respond to swipes or back gestures, with limits apps can tune.

  • Compose Remote: internals hidden from public APInew

    RCA-009: Hide creation and core internals from public API

    Creation and core internals leave the public surface, the kind of trimming done before a beta.

Show 15 more
  • WebView: share memory buffers with Javanew

    WebView Java SharedArrayBuffer API

    AndroidX WebKit is adding a Java API for SharedArrayBuffer, so apps could pass large binary data between native code and web content without copying.

  • CameraX: Ultra HDR sizes checked only on Android 15+new

    Only query JPEG_R extension sizes on API 35+ in Camera2ExtensionsVendorExtender

    JPEG_R is the Ultra HDR format; skipping the query on older versions avoids failures with vendor camera extensions.

  • Privacy Sandbox ads library removed

    Remove abandoned androidx.privacysandbox.ads library

    The abandoned Jetpack wrapper for Privacy Sandbox ads APIs is deleted, a sign that ads-side Privacy Sandbox work is no longer pursued in AndroidX.

  • App search: flag for an index flush fix

    Add flag enable_icing_term_index_flush_fix

    Icing is the engine behind AppSearch, Android's on-device search; the fix ships behind a flag so it can be rolled out gradually.

  • App search: flag for faster quantized scoring

    Add flag enableQuantizationScoringOptimization

    Quantized scoring suggests AppSearch is optimizing embedding or vector search on device, groundwork for semantic search in apps.

  • Compose PDF viewer gets an annotation toolbar

    Add AnnotationToolbar and AnnotationToolbarState in pdf-compose

    The Compose PDF library is moving from viewing toward marking up documents, likely paired with the Ink handwriting library.

  • Compose styles: properties added, kept internal

    Introduce style properties as internal

    The Compose styles API is growing behind internal visibility, so it can change freely before a public experimental release.

  • Build: old Gradle up-to-date workarounds removed

    Remove obsolete task up to date workarounds

    Build infrastructure cleanup in AndroidX; no API change for app developers.

  • Compose Remote: scope marker on layout builders

    Add RemoteLayoutScopeMarker to Remote layout scopes

    A DSL marker stops nested remote layouts from calling functions of an outer scope by mistake, API polish before a beta.

  • Internal modifier for one-handed gestures

    Add an internal oneHandedGesture modifier

    Likely Wear OS hand gestures, such as pinching fingers to act, coming to Compose as a modifier, kept internal for now.

  • Annotation processing: detect unreadable annotations

    Add XPropertyElement.hasInaccessibleAnnotations()

    XProcessing, the layer under Room and Hilt's processors, learns when annotations cannot be read, which should mean clearer errors instead of silent misses.

  • Internal settings for one-handed gestures

    Add an internal one-handed gesture configuration

    A configuration object to go with the one-handed gesture modifier, so apps could tune or disable gestures once it goes public.

  • Information Matrix: slot components added

    Add the Information Matrix slot components

    A new, unreleased component called Information Matrix is being built in stages; slot components suggest a grid of customizable data cells, possibly for watch or widget surfaces.

  • Information Matrix: slot contracts added

    Add the Information Matrix slot component contracts

    The second step of the Information Matrix component, defining what each slot accepts before the visible pieces land.

  • Information Matrix: contracts added

    Add the Information Matrix contracts

    The first building block of the unreleased Information Matrix component; its purpose is not yet public.

Experts

17Read Kevin Schildhorn's new part on hardening Turbine tests for Kotlin Flows, then skydoves' three-part Compose internals series and Chris Banes on Haze 2.0.
Show 9 more

Library releases

11KSP 2.3.13 adds an API for context parameters and is worth taking if your processors meet them, Roborazzi 1.77.0 fixes configuration cache reuse, Compose Unstyled 2.10.2 fixes dialogs inside a Host, Metro 1.5.1 and OkHttp 5.5.1 are safe patches, and Decompose 3.6.0 is still beta.
  • Compose Unstyled 2.10.2

    Patch Changes 649c471 : Make dialogs, modal bottom sheets, and modal drawers behave inside of a Host the same as without. Tab and Shift+Tab now stay inside the topmost modal, background controls receive no keystrokes, and closing the modal restores the previous focus. b59481b …

  • Metro 1.5.1

    Fixes [IR/IC] Avoid recompiling unrelated files in the graph module when the constructor of an upstream @Contributes* class changes. [Gradle] Fix false hidden-dependency reports for contributions from associated compilations, such as main sources in tests. This particularly…

  • Metro 1.5.0

    New [metrox-viewmodel-compose] Add support for linuxX64 . [FIR] Add the diagnostic-level compiler option to change the severity of Metro diagnostics reported in FIR. It works like kotlinc's -Xwarning-level but can also lower errors and use IDE-only levels. See changing…

  • OkHttp 5.5.1

    Version 5.5.1

  • OkHttp 4.12.1

    Version 4.12.1

  • KSP 2.3.13new

    What's Changed Add KSContextParameter and KSFunctionDeclaration.contextParameters to the API ( #3216 ) Improve diagnostics by returning last round dismissible errors when KSP fails ( #3196 ) Add opt-in processor classloader caching ( ksp.classloader.cache.processors=true ) to…

  • kotlinx.serialization 1.12.0

    This release is based on Kotlin 2.4.10 and contains all of the changes from 1.12.0-RC . Additionally, it includes a fix removing unnecessary leading whitespaces from Proguard rules ( #3270 ), and two changes stabilizing kotlin.uuid.Uuid serializer ( #3263 ) and…

  • Ktor 3.6.1

    Published 29 October 2026 Improvements KTOR-10005 Native accepted TCP sockets do not enable TCP_NODELAY, unlike JVM KTOR-9966 Netty poor performance on delay calls KTOR-4206 The "com.typesafe.config.ConfigException$Missing" exception leaks from the typesafe config library when…

Show 3 more
  • Apollo Kotlin 4.4.4

    A patch version to fix Java codegen generating bad code when using custom scalars mapped to Long or Float. 👷‍♂️ All changes [FIX] codegen: fix Java codegen generating bad code when using custom scalars mapped to Long or Float. ( #7032 )

  • Decompose 3.6.0-beta02

    What's Changed Library Add linuxX64 and linuxArm64 targets by @arkivanov-bot in #1028 Fix wrong directions in AndroidPredictiveBackAnimatableV1 in RTL layout by @arkivanov in #1030 Fix wrong directions in AndroidPredictiveBackAnimatableV2 in RTL layout by @arkivanov in #1031 Add…

  • Roborazzi 1.77.0

    Bugfix: verifyRoborazzi* and compareRoborazzi* reuse the configuration cache With the configuration cache enabled, the second verifyRoborazzi* or compareRoborazzi* run after clean recalculated the task graph because "the file system entry 'build/outputs/roborazzi' has been…

Leaks & ecosystem

22Google's numbers put Android 17 on about 1 percent of devices and Android 16 above 25 percent, Android Auto video apps are still coming soon, and teardowns point to an offline mode for Gemini Live.
Show 14 more

Community

37The malformed SMS that crashes Android's phone service is still drawing discussion, ProAndroidDev asks who debugs agent-written Kotlin at 2 AM, and Vela and Mochy Music trend among Kotlin repos.
Show 29 more
Start the Product & AI read4 stories, ranked by what you’d regret missing · 2 min read

  1. A backdoored Claude Code extension reached the Cursor marketplace

    Coding-agent extensions run with your full developer access, so where you install them from is now a security decision.

    Cursor published a security notice about a backdoored Claude Code extension for VS Code on its marketplace. It was the top Hacker News story, with about 810 points and 500 comments. Lobsters covered backdoors in Claude Code extensions for VS Code and Cursor, and a Dev.to audit says a trojanized build on Open VSX planted an SSH backdoor as soon as the editor loaded it.

    HN front page · Across sourcesRead the sourceDiscussion

  2. A working Nintendo built from about one gigabyte of Qwen weights

    Models that imitate whole machines are moving from lab papers to hobby projects anyone can download and try.

    A Show HN post claims a working Nintendo console that runs inside about one gigabyte of weights from Qwen, Alibaba's open model family. The idea is that a trained model behaves like the console, rather than an emulator written by hand. It led Show HN and reached the front page, with about 200 points and over 100 comments.

    Show HN · In 2 sourcesRead the sourceDiscussionProduct notes

  3. Essay argues AI is deskilling site reliability engineers

    Teams handing on-call work to AI should decide which skills people must keep practising by hand.

    An essay titled AI isn't replacing SREs, it's deskilling them reached number two on Hacker News, with about 530 points and nearly 400 comments. SREs are the engineers who keep production systems running. Going by the title, the argument is that relying on AI tools wears down the skills people need when those tools fail.

    HN front page · Worth a debateRead the sourceDiscussion

  4. DeepTutor, an open-source personal tutor, adds over 2,000 stars

    Tutoring is one of the clearest consumer uses for agents, and an open-source base lets schools and builders run their own.

    DeepTutor, from the HKUDS research group, is an open-source project for lifelong personalized tutoring. It gained about 2,300 stars in a day and is near 52 thousand, among the fastest climbers on GitHub today. It arrives alongside other learning tools, from a Y Combinator company building AI-native schools to Gemini Notebook adding textbooks.

    GitHub Trending · Fastest risingRead the sourceProduct notes

Go deeper: themes, product notes, every source

Themes across sources

6Agent supply-chain safety · Coding-agent control surfaces · Anti-slop design skills · Agent memory layers · World models · On-device AI

Ideas showing up in more than one place today. The line counts items per day over two weeks.

Agent supply-chain safety

Rising · 7 items from 3 sources

The top Hacker News story, with about 810 points and 500 comments, is a backdoored Claude Code extension for VS Code on the Cursor marketplace. Lobsters covered backdoors in Claude Code extensions for VS Code and Cursor, and a Dev.to audit says a trojanized build on Open VSX planted an SSH backdoor when the editor loaded it. Around it, posts dissect the eMudhra certificate compromise for pip and C2PA, a false typo-squat flag on npm, and defending the patch gap against automated attackers.

Why now. Coding agents now ship as editor extensions with shell access, so one poisoned marketplace listing reaches a developer's keys and repositories directly.

Backdoor in Claude Code for VS Code extension on the Cursor MarketplaceHN front page

Coding-agent control surfaces

Rising · 11 items from 3 sources

GitHub keeps filling with places for agents to work: CopilotKit's OpenBot gives each agent its own computer and records every action, OpenMausBot ships a virtual machine for bots, and t3code and JetBrains' Thinkrail each add hundreds of stars. Product Hunt adds MAO for running Claude Code, Codex and Grok on one canvas, ilse for pointing at your app to change its code, Faylee for bug reports an agent can fix, and Flatbread for sharing a whole agent run. On Android, a ProAndroidDev post runs DeepSeek, MiniMax and GLM inside Claude Code and Android Studio.

Why now. Teams run several agents at once, so the open questions are where agents live, what they can touch and how people review their work.

CopilotKit/OpenBotGitHub Trending

More on this

Anti-slop design skills

Rising · 6 items from 2 sources

Impeccable added about 2,300 stars today and is near 97 thousand, while Superpowers passed 309 thousand and Matt Pocock's skills collection is near 308 thousand. Anthropic's public skills repository and a new Expo skills pack for React Native projects also trend, and a Dev.to post describes building an AI to catch bad web design patterns.

Why now. Skills are a shared format across most coding agents, so one file of rules now reaches every agent a team uses.

pbakaus/impeccableGitHub Trending

Agent memory layers

Rising · 7 items from 4 sources

claude-mem, which records what an agent did and feeds the useful parts back next session, leads GitHub with about 1,140 new stars and is past 106 thousand. Macro pitches a team workspace with shared AI memory, a paper on a sleep framework for continual learning reached the Hacker News front page, a teardown shows Google building a page for everything Gemini remembers, and Product Hunt lists Recall for Mac, Smart Jelly and Kin.

Why now. Agents now run long, repeated tasks, and starting every session from zero has become the obvious waste to remove.

thedotmack/claude-memGitHub Trending

World models

Steady · 4 items from 4 sources

A working Nintendo in about 1 GB of Qwen weights led Show HN and reached the front page, a day after a developer turned DOOM into a transformer. Hugging Face papers add L3G-World, which generates endless explorable worlds in real time, and Product Hunt lists Viggle JST-2 for 3D scenes and motion from one model.

Why now. Open models are small and cheap enough to fine-tune at home, so making a network imitate a whole machine or game is now a weekend project.

Show HN: A working Nintendo in ~1GB of Qwen weightsShow HN

On-device AI

Rising · 9 items from 5 sources

Teardowns show Google working on an offline mode for Gemini Live, covered by both Android Authority and its Insights team. Liquid AI's LFM3-1.7B and LFM3-Omni-2B, NVIDIA's 4 GB Nemotron 3.5 Bolt and the 4B Spark X2.6 trend on Hugging Face, and Android Weekly covers Gemini Nano's input limit, shipping a 584 MB Gemma model through Play and on-device agents with Google's ADK.

Why now. Small open models now handle everyday tasks, and Google's own assistant is moving toward running without a network.

No internet, no problem: Google is working on an offline mode for Gemini LiveAuthority Insights (leaks & teardowns)

Fading: Verifying agent work (last seen 2026-10-30) · Open-weight frontier models (last seen 2026-10-30) · Agent-ready Android (last seen 2026-10-29) · Android XR glasses (last seen 2026-10-29) · Agents as security auditors (last seen 2026-10-26) · Open voice models (last seen 2026-10-26) · Agent sandboxes (last seen 2026-10-25)

Product notes

15Who each product is for, the problem, why now, its wedge and how it makes money. Today’s picks first.

Plain text: the source says soDotted: inferred

Show HN: A working Nintendo in ~1GB of Qwen weights

productShow HN, HN front page206 pts · 116 commentsHype flag

What
A Show HN project that claims a working Nintendo console running inside about 1 GB of Qwen model weights.
For whom
Researchers and hobbyists curious whether a language model can imitate a whole machine.
Problem
Emulators are written by hand instruction by instruction; this asks whether a network can learn the behaviour instead.
Why now
Small open models like Qwen are cheap to fine-tune at home, and DOOM-as-a-transformer style projects showed the idea works for games.
Wedge
A famous, nostalgic machine as a demo people want to try.
Money
Hobby project, no model.
Like
GameNGen · DOOM transformer · Genie 4 · Oasis

A demo with no product; the brief does not say how complete, accurate or fast the imitation is.

LessonA recognisable demo spreads a research idea further than a paper; pick the example people already love.

HKUDS/DeepTutor

repoGitHub Trending★ 2,282 today

What
DeepTutor, an open-source project for lifelong personalized tutoring from the HKUDS group.
For whom
Learners who want a tutor that remembers what they know, and builders who want a base for their own tutoring product.
Problem
General chatbots answer questions but do not track a learner's progress; people cope with separate notes and flashcard apps.
Why now
Models can now explain at many levels and agents can keep long-term memory of a learner.
Wedge
Personalization over time rather than single answers.
Money
Open source from a research group; a hosted site exists, no stated pricing.
Like
Khanmigo · NotebookLM · Gemini Notebook · Duolingo Max

LessonMemory turns a question-answer tool into a relationship; the second session is where tutoring products win or lose.

pbakaus/impeccable

repoGitHub Trending★ 2,300 today

What
Impeccable, a design vocabulary for coding agents: steer with commands like polish and bolder, review with critique and audit, and catch generic AI-looking design before it ships.
For whom
Developers who build front ends with coding agents and keep getting the same bland layouts.
Problem
Agent-made UIs look alike and need many rounds of vague feedback; people cope by pasting screenshots and style notes into every prompt.
Why now
Coding agents now load shared skill files, so design rules written once travel with every task.
Wedge
A small set of named verbs that make design feedback repeatable.
Money
Open source, no model yet.
Like
Superpowers · Anthropic skills · v0 · Figma Make

LessonGive users words for what they already feel; a shared vocabulary is easier to adopt than a new tool.

thedotmack/claude-mem

repoGitHub Trending★ 1,140 today

What
claude-mem captures what an agent does in a session, compresses it with AI and injects the relevant parts into future sessions; it works with Claude Code, Codex, Gemini, Copilot, OpenCode and others.
For whom
Developers who use coding agents daily on the same projects.
Problem
Each new session forgets earlier decisions, so people keep long notes files or re-explain the project.
Why now
Agent hooks and long sessions are common now, so there is a record worth saving and a place to inject it.
Wedge
Works across many agents instead of locking memory to one vendor.
Money
Open source, no model yet.
Like
Mem0 · Letta · Zep · CLAUDE.md files

LessonSit between tools that do not talk to each other; neutral glue gets adopted faster than another walled garden.

CopilotKit/OpenBot

repoGitHub Trending★ 706 today

What
OpenBot from CopilotKit: open-source AI coworkers that each get their own computer with a browser, files and tools, with every action decided before it happens and recorded after.
For whom
Teams that want agents to do real desk work but need to audit what they did.
Problem
Agents that act on a shared machine are hard to contain and harder to review; teams cope with screen recordings and manual checks.
Why now
Computer-use models are good enough to finish tasks, so isolation and audit trails have become the blocker.
Wedge
An action log decided before and recorded after each step.
Money
Open source; CopilotKit likely earns from hosted or enterprise versions.
Like
OpenMausBot · Anthropic computer use · OpenAI Operator · E2B

LessonWhen a capability becomes common, sell the trust around it: logs, limits and a way to check the work.

lightpanda-io/browser

repoGitHub Trending★ 626 today

What
Lightpanda, a headless browser built for AI agents and automation rather than adapted from a desktop browser.
For whom
Teams running scrapers, test suites or browser agents at scale.
Problem
Headless Chrome is heavy on memory and slow to start, so large agent fleets cost a lot to run.
Why now
Browser agents multiplied the number of automated sessions, making browser cost per task a real line item.
Wedge
A browser for machines only, with no rendering for humans to pay for.
Money
Open source with a likely hosted cloud offering.
Like
Headless Chrome · Playwright · Browserbase · fastbrowse

LessonWhen a new kind of user appears, rebuilding a classic tool just for them can beat adapting the old one.

Faylee

productProduct Hunt#15 on Product Hunt

What
Faylee turns a bug report into one link that a coding agent can use to fix the bug.
For whom
Product teams where testers or support staff report bugs and developers fix them with agents.
Problem
Bug reports miss logs, steps and context, so developers re-ask questions or reproduce by hand before an agent can help.
Why now
Agents can fix many bugs on their own if they get full context, so the report itself becomes the input that matters.
Wedge
A single shareable link instead of a new tracker.
Money
Likely per-seat subscription.
Like
Jam · Bird Eats Bug · Marker.io · Sentry

LessonWhen the consumer of your output changes from a person to an agent, redesign the output, not just the workflow.

ilse

productProduct Hunt#13 on Product Hunt

What
ilse lets you point at part of your running app so your AI agent changes the matching code.
For whom
Designers and developers making visual tweaks who find it slow to describe a UI element in words.
Problem
Telling an agent which button or card to change takes long prompts and file hunting.
Why now
Agents can now edit code reliably; locating the right code from the screen is the remaining friction.
Wedge
Pointing instead of describing.
Money
Likely subscription; not stated.
Like
Onlook · Stagewise · Builder.io Visual Copilot · React DevTools

LessonThe fastest interface is often a gesture people already use; pointing beats prompting for visual work.

Canva Brush

productProduct Hunt#5 on Product Hunt

What
Canva Brush, a vision-language model inside Canva that edits your designs.
For whom
Canva users who make social posts, decks and marketing graphics.
Problem
Small edits like moving, restyling or replacing elements take many clicks; people cope with templates.
Why now
Models can now see a layout and edit its parts, not just generate a new image.
Wedge
Editing designs users already have, inside the tool they already use.
Money
Part of Canva's paid plans.
Like
Adobe Firefly · Figma AI · Photoshop Generative Fill

LessonIncumbents win AI features by applying them to the user's own files, which a new entrant does not have.

Mirrors: Fix recurring AI agent mistakes with real-time guidance

productYC Launches15 votes

What
Mirrors, a Y Combinator company, fixes recurring AI agent mistakes with real-time guidance.
For whom
Teams running agents in production that keep failing in the same ways.
Problem
Teams play whack-a-mole with agent errors by editing prompts after each failure.
Why now
Agents are in production long enough for repeated failure patterns to show up in logs.
Wedge
Guidance at run time rather than retraining or rewriting prompts.
Money
Likely usage-based SaaS.
Like
Langfuse · Braintrust · Arize · Guardrails AI

LessonName the user's coping habit in your pitch; 'whack-a-mole' tells buyers you know their week.

LiquidAI/LFM3-1.7B

modelHF trending modelstrending 552

What
LFM3-1.7B, a 1.7 billion parameter model from Liquid AI, third on Hugging Face's trending list.
For whom
App developers who want a capable model running on phones and laptops.
Problem
Cloud models add cost, delay and privacy questions; small models used to be too weak for real tasks.
Why now
Small models have improved enough, and phone runtimes like LiteRT have matured, to make on-device use practical.
Wedge
Model sizes picked for edge devices first.
Money
Open weights with licensing and enterprise deals.
Like
Gemma · Nemotron 3.5 Bolt · Qwen small models · Phi

LessonChoose a constraint, such as fitting on a phone, and win it outright rather than ranking tenth overall.

Durable Actors: The perfect primitive for Multiplayer AI and Agent Coordination

productYC Launches4 votes

What
Durable Actors, an open-source alternative to Cloudflare's Durable Objects, pitched as the primitive for multiplayer AI and agent coordination.
For whom
Developers building collaborative apps or many agents that share state.
Problem
Coordinating shared state between users and agents needs locks, queues and databases glued together.
Why now
Multi-agent apps need one owner per piece of state, which is what the actor model gives.
Wedge
An open version of a primitive that is otherwise tied to one cloud.
Money
Open source with a likely hosted service.
Like
Cloudflare Durable Objects · Rivet · Orleans · Akka

LessonOpen-sourcing a proprietary cloud primitive gives buyers an exit, which is often why they say yes.

Haze 2.0

articleChris BanesAndroid

What
Haze 2.0, Chris Banes' blur and glass effects library for Compose Multiplatform, adds Glass, a new effects API, performance improvements and experimental Android native backdrop support.
For whom
Compose and Compose Multiplatform developers who want frosted-glass bars and sheets that match iOS styles.
Problem
Blurring what sits behind a component is hard to do fast on Android, so teams fake it with flat translucent colors.
Why now
Glass looks are back in fashion with Apple's Liquid Glass, and Android now offers platform support for backdrop effects.
Wedge
One API across Android, iOS and desktop.
Money
Open source, no model.
Like
Cloudy · RenderEffect · UIVisualEffectView

LessonFollow design trends with infrastructure, not screens; the library everyone needs for the trend outlasts it.

Appfunctions 1.0.0-beta01

releaseAndroidX releasesAndroid

What
AppFunctions 1.0.0-beta01, the Jetpack library that lets an app expose actions which assistants and agents on the device can call.
For whom
Android apps with tasks users would ask an assistant to do, such as booking, adding notes or starting a workout.
Problem
Assistants can only open apps or rely on deep links, so they cannot complete tasks inside third-party apps.
Why now
On-device and Gemini agents need typed, permissioned entry points into apps, much like MCP servers on the web.
Wedge
A Jetpack library app teams can add without platform changes.
Money
Free platform API; Google gains by making Android agent-ready.
Like
App Intents (iOS) · MCP servers · App Actions · Shortcuts

LessonBeta is the moment to shape an API; teams that try it now get their use case designed in.

PimpinPumpkin/Vela

repoGitHub Trending · KotlinAndroid★ 16 today

What
Vela, maps and turn-by-turn navigation for Android without Google Play Services, built on open map data with Google lookups only if you choose them.
For whom
People on de-Googled phones and privacy-minded users who still want usable navigation.
Problem
Navigation apps without Play Services are often clunky, so users fall back to Google Maps or go without.
Why now
Open map data and routing engines have matured enough to rival commercial maps for daily driving.
Wedge
Privacy by default with Google lookups as an opt-in.
Money
Open source, no model yet.
Like
Organic Maps · OsmAnd · CoMaps · Google Maps

LessonMake the trade-off optional instead of absolute; letting users opt into the convenient service widens your audience.

By source

11Everything collected today from GitHub Trending, Product Hunt, Show HN, HN front page, YC Launches, HF trending models, HF trending spaces, HF daily papers, Lobsters, Dev.to top, Techmeme, in each source’s own order.

GitHub Trending

15
  • thedotmack/claude-memcard

    Persistent Context Across Sessions for Every Agent – Captures everything your agent does during sessions, compresses it with AI, and injects relevant context back into future sessions. Works with Claude Code, OpenClaw, Codex, Gemini, Hermes, Copilot, OpenCode + More

    ★ 1,140 today

  • milind-soni/OpenMausBot

    OpenMausBot, an open-source bot with its own virtual machine, adds about 900 stars.

    ★ 895 today

  • pbakaus/impeccablecard

    The design vocabulary for your coding agent. Steer with polish and bolder, review with critique and audit, catch AI slop before it ships.

    ★ 2,300 today

  • CopilotKit/OpenBotcard

    Open-source AI coworkers that each get a computer of their own: a browser, files and tools, with every action decided before it happens and recorded after. Bring any AG-UI agent.

    ★ 706 today

  • Renhuai123/ziwei-doushunew

    An open-source Zi Wei Dou Shu astrology chart engine with classical texts, among today's fastest risers.

    ★ 1,126 today

  • expo/skillsnew

    Expo publishes agent skills for Expo and EAS, so coding agents follow React Native project conventions.

    ★ 38 today

  • macro-inc/macronew

    Macro links email, chat, docs, tasks and CRM in one workspace with shared AI memory.

    ★ 57 today

  • mattpocock/skillsnew

    Matt Pocock's skills collection adds about 880 stars, near 308 thousand.

    ★ 876 today

Show 7 more
  • pingdotgg/t3codenew

    t3code adds about 930 stars, one of the busiest coding-agent repos today.

    ★ 930 today

  • HKUDS/DeepTutorcard

    DeepTutor: Lifelong Personalized Tutoring. https://deeptutor.info/ .

    ★ 2,282 today

  • anthropics/skillsnew

    Anthropic's public Agent Skills repository passes 185 thousand stars.

    ★ 295 today

  • JetBrains/thinkrail

    JetBrains' Thinkrail, a lightweight IDE built around a coding agent, adds about 720 stars.

    ★ 722 today

  • togg53192-cmd/jailbreaks

    LIST OF ALL MY JAILBREAKS

    ★ 2,055 today

  • obra/superpowers

    Superpowers, the agentic skills framework, passes 309 thousand stars.

    ★ 1,259 today

  • lightpanda-io/browsernewcard

    Lightpanda: the headless browser designed for AI and automation

    ★ 626 today

Product Hunt

50
  • StereoLift Livenew

    StereoLift Live shows your desktop in 3D on a Quest headset, the top Product Hunt launch.

    #1 on Product Hunt

  • Vaani Developer Platformnew

    Vaani pitches a developer platform for building voice agents.

    #2 on Product Hunt

  • Kestrixnew

    Kestrix runs CI across repositories with dependency-graph triggers.

    #3 on Product Hunt

  • MCP Server by XWikinew

    XWiki's MCP server gives AI assistants access to a wiki while keeping permissions in place.

    #4 on Product Hunt

  • Canva Brushnewcard

    A vision-language model that edits your designs

    #5 on Product Hunt

  • Knocket AInew

    A chat widget for talking to website visitors.

    #6 on Product Hunt

  • SpeakON Agentnew

    SpeakON turns spoken requests into tasks for an agent.

    #7 on Product Hunt

  • Cluinnnew

    Cluinn answers questions during a call on a Mac, not after it.

    #8 on Product Hunt

Show 42 more
  • sitesolidenew

    Deploy vibe-coded apps on your own server instead of a hosted platform.

    #9 on Product Hunt

  • StepGrabnew

    StepGrab's step-by-step guide maker reaches Windows with redaction.

    #10 on Product Hunt

  • Axonnicnew

    Axonnic's sleeves sense muscle activity to record hand data for robot training, also a YC launch.

    #11 on Product Hunt

  • Norah Studionew

    A new app for exploring your music library.

    #12 on Product Hunt

  • ilsenewcard

    Point at your app and your AI agent changes the code

    #13 on Product Hunt

  • SparkBots AI (Powered by Jev)new

    Extracts key details from any webpage or PDF.

    #14 on Product Hunt

  • Fayleenewcard

    Bug reports your coding agent can fix from one link

    #15 on Product Hunt

  • Noctonew

    Another app that puts tools in the MacBook notch.

    #16 on Product Hunt

  • iSurgenew

    An activity monitor, per-app volume and disk cleanup in the Mac notch.

    #17 on Product Hunt

  • Teable Mobilenew

    Teable's no-code database builder comes to phones.

    #18 on Product Hunt

  • fastbrowsenew

    fastbrowse claims to be the fastest open-source browser agent.

    #19 on Product Hunt

  • Chopsticknew

    Mac widgets that write themselves from your activity.

    #20 on Product Hunt

  • Gemini 4 Carbon

    Gemini 4 Carbon, Google's latest frontier model, stays on Product Hunt.

    #21 on Product Hunt

  • Voxel51 Multimodal

    Voxel51 extends its data tooling to multimodal robot data.

    #22 on Product Hunt

  • OutreachGuy

    An AI sales rep that follows up on leads you dropped.

    #23 on Product Hunt

  • Confetti

    Confetti designs hardware gadgets from plain English.

    #24 on Product Hunt

  • Intrascope 2.0

    Intrascope 2.0 adds an agent, 1,650 app connections and model routing.

    #25 on Product Hunt

  • Arco

    GitHub and Linear issues inside Codex, so the agent sees the ticket.

    #26 on Product Hunt

  • AfterLid

    Keeps a Mac working with the lid closed and lets you check in from a phone, built for long agent runs.

    #27 on Product Hunt

  • ChatInbox

    An inbox grouped by contact instead of by thread.

    #28 on Product Hunt

  • StoreLocale

    Translates the app UI inside App Store and Google Play screenshots, a store-listing chore most teams skip.

    #29 on Product Hunt

  • AutoShelf 2.6

    AutoShelf 2.6 automates Mac folders with if-this-then-that rules.

    #30 on Product Hunt

  • Kin

    Kin pitches user-owned memory so AI labs answer to the user.

    #31 on Product Hunt

  • Floot in ChatGPT

    Build and publish apps without leaving ChatGPT.

    #32 on Product Hunt

  • MAO

    MAO puts Claude Code, Codex and Grok on one shared canvas.

    #33 on Product Hunt

  • Viggle JST-2

    Viggle JST-2 makes 3D scenes, characters and motion from one model.

    #34 on Product Hunt

  • Suno MCP

    Suno's music generation exposed as an MCP server for agents.

    #35 on Product Hunt

  • Gologin Cloud Browser

    A cloud browser for agents that need to get through difficult sites.

    #36 on Product Hunt

  • TopAgent

    An agent that drives attendance for B2B events.

    #37 on Product Hunt

  • Reset Alerts

    Text alerts when your Codex usage resets, so paid quota is not wasted.

    #38 on Product Hunt

  • Radar

    Radar gives your existing agent production context instead of selling a separate AI SRE.

    #39 on Product Hunt

  • IFTTT.ai

    IFTTT relaunches around AI automation.

    #40 on Product Hunt

  • Empacer

    A voice-first assistant for small business owners.

    #41 on Product Hunt

  • Scape

    An email inbox that sorts and acts on mail for you.

    #42 on Product Hunt

  • Polylane for ChatGPT

    Ask ChatGPT what broke in production, incident data inside the chat app.

    #43 on Product Hunt

  • Flatbread

    Flatbread shares a whole agent run instead of a screenshot.

    #44 on Product Hunt

  • AssemblyAI

    AssemblyAI adds text-to-speech built for voice agents.

    #45 on Product Hunt

  • Enclave

    Enclave puts the context you need behind a single hover.

    #46 on Product Hunt

  • JIMMY

    JIMMY, an agent that runs on your PC.

    #47 on Product Hunt

  • Notch Metrics

    Build analytics widgets for the Mac notch with Claude Code.

    #48 on Product Hunt

  • Smart Jelly

    Everyday AI tools and memory in one Mac app.

    #49 on Product Hunt

  • Recall for Mac

    Recall for Mac records what happens on your Mac so you can search it later.

    #50 on Product Hunt

Show HN

12
Show 4 more

HN front page

30
Show 22 more

YC Launches

20
Show 12 more

HF trending models

19
Show 11 more

HF trending spaces

17
Show 9 more

HF daily papers

20
Show 12 more

Lobsters

25
Show 17 more

Dev.to top

20
Show 12 more

Techmeme

15
Show 7 more

About this issue

406 items from 77 sources, 9 of them seen in several sources. 5 sources failed to fetch.

Source health

77Which sources were fetched today, and how many items each returned.
SourceSectionItemsStatus
GitHub TrendingMain15Fetched
Product HuntMain50Fetched
Show HNMain12Fetched
HN front pageMain30Fetched
YC LaunchesMain20Fetched
HF trending modelsMain20Fetched
HF trending spacesMain20Fetched
HF daily papersMain20Fetched
LobstersMain25Fetched
Dev.to topMain20Fetched
TechmemeMain15Fetched
AndroidX releasesAndroid · Official & releases58Fetched
Android Developers BlogAndroid · Official & releases5Fetched
Android Studio releasesAndroid · Official & releases5Fetched
Kotlin blogAndroid · Official & releases1Fetched
Kotlin releasesAndroid · Official & releases0Fetched
Compose Multiplatform releasesAndroid · Official & releases1Fetched
Android Developers YouTubeAndroid · Official & releases0Failed
Kotlin on YouTube (JetBrains)Android · Official & releases0Failed
Google Developers BlogAndroid · Official & releases1Fetched
AndroidX merged API changesAndroid · AOSP / AndroidX insider23Fetched
Dove Letter (skydoves)Android · Experts5Fetched
skydoves on MediumAndroid · Experts0Fetched
Jake WhartonAndroid · Experts0Fetched
Zac SweersAndroid · Experts0Fetched
Chris BanesAndroid · Experts2Fetched
Ryan HarterAndroid · Experts0Fetched
Talking KotlinAndroid · Experts0Fetched
Joe BirchAndroid · Experts2Fetched
CommonsWare (Mark Murphy)Android · Experts3Fetched
TechYourChance (Vasiliy Zukanov)Android · Experts0Failed
Kt. Academy (Marcin Moskała)Android · Experts2Fetched
Touchlab (Kotlin Multiplatform)Android · Experts3Fetched
Landscapist releasesAndroid · Library releases0Fetched
Balloon releasesAndroid · Library releases0Fetched
Sandwich releasesAndroid · Library releases0Fetched
Cloudy releasesAndroid · Library releases0Fetched
Compose Stability Analyzer releasesAndroid · Library releases0Fetched
Haze releasesAndroid · Library releases0Fetched
Compose Unstyled releasesAndroid · Library releases1Fetched
Coil releasesAndroid · Library releases0Fetched
Circuit releasesAndroid · Library releases0Fetched
Metro releasesAndroid · Library releases2Fetched
Molecule releasesAndroid · Library releases0Fetched
Turbine releasesAndroid · Library releases0Fetched
Paparazzi releasesAndroid · Library releases0Fetched
SQLDelight releasesAndroid · Library releases0Fetched
OkHttp releasesAndroid · Library releases2Fetched
Retrofit releasesAndroid · Library releases0Fetched
LeakCanary releasesAndroid · Library releases0Fetched
Koin releasesAndroid · Library releases0Fetched
Dagger / Hilt releasesAndroid · Library releases0Fetched
KSP releasesAndroid · Library releases1Fetched
kotlinx.coroutines releasesAndroid · Library releases0Fetched
kotlinx.serialization releasesAndroid · Library releases1Fetched
Ktor releasesAndroid · Library releases1Fetched
Apollo Kotlin releasesAndroid · Library releases1Fetched
Kermit releasesAndroid · Library releases0Fetched
Store releasesAndroid · Library releases0Fetched
Decompose releasesAndroid · Library releases1Fetched
Orbit MVI releasesAndroid · Library releases0Fetched
Roborazzi releasesAndroid · Library releases1Fetched
detekt releasesAndroid · Library releases0Fetched
ktlint releasesAndroid · Library releases0Failed
MockK releasesAndroid · Library releases0Fetched
Lottie releasesAndroid · Library releases0Fetched
Authority Insights (leaks & teardowns)Android · Leaks & ecosystem8Fetched
9to5Google · AndroidAndroid · Leaks & ecosystem8Fetched
Android AuthorityAndroid · Leaks & ecosystem6Fetched
Android FaithfulAndroid · Leaks & ecosystem1Fetched
Android WeeklyAndroid · Community15Fetched
GitHub Trending · KotlinAndroid · Community13Fetched
Lobsters · androidAndroid · Community1Fetched
HN · AndroidAndroid · Community6Fetched
ProAndroidDevAndroid · Community5Fetched
Kotlin WeeklyAndroid · Community1Fetched
Philipp Lackner (YouTube)Android · Community0Failed
1 of 5

Next Android release ends direct contact writes for new apps

If your app adds or edits contacts, raising your target SDK will break that code unless you move to the new approach first.

Google's Android Developers Blog says apps that target the upcoming Android release will no longer be able to write to contacts directly with the WRITE_CONTACTS permission. The post, titled a more reliable way to add and edit contacts, describes the path apps should use instead. Apps that create, import or sync contacts need to move before they raise their target SDK.

Android Developers Blog · OfficialRead the source

1 of 4

A backdoored Claude Code extension reached the Cursor marketplace

Coding-agent extensions run with your full developer access, so where you install them from is now a security decision.

Cursor published a security notice about a backdoored Claude Code extension for VS Code on its marketplace. It was the top Hacker News story, with about 810 points and 500 comments. Lobsters covered backdoors in Claude Code extensions for VS Code and Cursor, and a Dev.to audit says a trojanized build on Open VSX planted an SSH backdoor as soon as the editor loaded it.

HN front page · Across sourcesRead the sourceDiscussion