Android Feed

← The 5 stories that mattered

Everything collected ·

All 388 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. Grab cut ANRs by a third with R8 and app profiles

    If your app still uses old ProGuard rules or has no Baseline Profile, this is a measured case for fixing both.

    The Android Developers Blog describes how Grab, the Southeast Asian superapp, tuned its build. It first switched on R8 optimization, then moved to R8 full mode, and added custom Baseline and Startup Profiles for its startup paths. Overall, freezes known as ANRs fell about 32 percent, cold start improved about 35 percent, and the app shrank about 30 percent.

    Android Developers Blog · OfficialRead the source

  2. New MANAGE_CONTACTS access replaces direct contact writes

    Contact-saving code needs a plan now, because the right path depends on whether your app qualifies for the new access.

    Google detailed the contacts change. Apps that target the 2027 Android release cannot write contacts with WRITE_CONTACTS. Most apps should open the system editor with insert or edit intents, which need no permission. Apps whose main job is managing contacts, such as backup or caller ID apps, can ask for a new special access called MANAGE_CONTACTS, arriving in Android 17 QPR2 with a Play declaration.

    Android Developers Blog · OfficialRead the source

  3. Glance adaptive widgets take shape in AndroidX code

    Teams maintaining several widget sizes may soon replace them with a single adaptive Glance widget.

    New changes in AndroidX, Android's set of support libraries, add a receiver for adaptive app widgets in Glance, the library for building widgets with Compose-style code. Related changes prepare adaptive templates for public API and give Wear widgets a profile builder with padding and corner radius. Together they point to widgets that fit many sizes from one layout.

    AndroidX merged API changes · AOSP insiderRead the source

  4. Did AI really change native versus cross-platform?

    If AI makes native code cheap, the main argument for cross-platform frameworks weakens, which affects how teams plan Android and iOS work.

    Shopify said it is leaving React Native because AI coding tools changed the trade-off between native and cross-platform apps. Vasiliy Zukanov of TechYourChance reviewed the evidence behind that claim. The post asks whether writing the same feature twice really got cheap enough with AI help to beat sharing code across platforms.

    TechYourChance (Vasiliy Zukanov) · ExpertRead the source

  5. Gemini may soon set up your eSIM

    System tasks moving behind assistant connectors is a hint that apps will be expected to expose actions to Gemini too.

    An Android Authority teardown of the Google app found a new Gemini connector for eSIMs, the digital SIM cards built into phones. It can already start the eSIM setup process from a few simple prompts. This is a leak, so details and timing may change, but it follows other connectors that let Gemini act inside apps and settings.

    Android Authority · EcosystemRead the source

Also worth knowing

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

  • Android Studio Rabbit 2 RC 2 now available

    Android Studio Rabbit 2 reached RC 2 in the Beta channel, and Rabbit 3 is in Canary.

    Teams planning the next IDE upgrade; stable Rabbit 2 is close.

    WatchOfficial

  • 1.13.0-beta01

    Compose Multiplatform 1.13.0-beta01 is out; desktop libraries using the experimental DragAndDropTransferAction must be recompiled because of a binary-incompatible change.

    Library authors shipping Compose Multiplatform desktop code with drag and drop.

    WatchOfficial

Go deeper: every Android item, by lane

Official & releases

51No new Jetpack wave today; the news is policy and practice, as apps targeting the 2027 release must add contacts through system intents or request the new MANAGE_CONTACTS access, and Grab shows what R8 full mode and Baseline Profiles buy a large app.

AndroidX library releases

Stable · safe to ship · 8

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

Alpha · experimental, APIs may change · 20
Show 2 more

AOSP / AndroidX insider

27Gerrit shows Glance adaptive widgets and a Wear widget profile builder heading toward public API, Compose Remote gaining right-to-left layout, and Ink opening its animation APIs.
Show 19 more

Experts

19Read TechYourChance on whether AI really changed the native versus cross-platform trade-off, Joe Birch on pixelating paid content in Compose Multiplatform, and Chris Banes on Haze 2.0 next to Compose's new progressive blur.
Show 11 more

Library releases

7kotlinx.serialization 1.12.0 and Metro 1.5.0 are safe updates, OkHttp 5.5.1 and 4.12.1 are patch releases worth taking given this week's OkHttp CVE write-up, Roborazzi 1.77.0 fixes configuration cache reuse, and Decompose 3.6.0 is still in beta.
  • Compose Unstyled 2.10.2new

    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 1.5.0

    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

  • 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…

  • 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

21Teardowns point to Gemini setting up eSIMs, a new look for Gemini Live, Google Wallet payments while the phone stays locked, and Find Hub alert fixes ahead of Pixel Tag.
Show 13 more

Community

41Hacker News is busy with iOS firmware on Android devices and a TCL Android TV root, while Android Weekly covers AppFunctions, Gemini Nano's token limit and smoother Compose screens.
Show 33 more
Start the Product & AI read4 stories, ranked by what you’d regret missing · 2 min read

  1. DeepMind's Genie 4 pushes world simulation forward

    Generated worlds are becoming a cheap place to train and test agents and robots before they touch the real one.

    Google DeepMind announced Genie 4, which its blog calls a new generation of world simulation. World models generate an environment you can move through and act in, frame by frame, rather than a fixed video. It was one of the biggest Hacker News stories of the day, with about 790 points and nearly 500 comments. Several top Hugging Face papers also studied world models for robots.

    HN front page · Worth a debateRead the sourceDiscussionProduct notes

  2. Liquid AI's LFM3 brings many input types to tiny models

    Small models that handle more than text make on-device features possible without a cloud call.

    Liquid AI introduced LFM3, a model family it describes as every modality for specialized intelligence. The announcement reached the Hacker News front page with about 130 points, and its 1.7 billion parameter model is trending on Hugging Face. Models this size can run on phones and small devices, aimed at focused tasks rather than general chat.

    HF trending models · Fastest risingRead the sourceProduct notes

  3. Python's experimental JIT is set to be removed

    Teams hoping for a built-in speed-up in standard Python should not plan around the JIT for now.

    LWN reports that the experimental just-in-time compiler in CPython, the main Python interpreter, is set to be removed. The JIT first shipped as an opt-in experiment a few releases ago. The news drew about 465 points and 250 comments on Hacker News and was a top story on Lobsters.

    HN front page · In 2 sourcesRead the sourceDiscussion

  4. Hallmark keeps leading GitHub as design skills grow

    Design rules shipped as a text file now reach every coding agent a team uses.

    Hallmark, an open-source design skill that steers Claude Code, Cursor and Codex away from the generic AI look, added about 2,800 stars today, more than any other repository. It now has about 39 thousand. Impeccable, a similar design vocabulary for agents, passed 90 thousand, and a new collection of AI design skills is near 10 thousand.

    GitHub Trending · Fastest risingRead the sourceProduct notes

Go deeper: themes, product notes, every source

Themes across sources

6World models · Coding-agent control surfaces · On-device AI · Anti-slop design skills · Verifying agent work · Agent-ready Android

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

World models

New · 6 items from 4 sources

Google DeepMind's Genie 4, billed as a new generation of world simulation, drew nearly 500 comments on Hacker News. The same day, the top Hugging Face papers include Hy-Nomas for interactive world generation, Rollback for improving robot policies inside a world model, and ThinkWorld for imagined dynamics in robot reasoning. A YC launch, Hemlock, sells force data for robot training.

Why now. Video models now hold a scene together over many steps, so labs use them as cheap simulators for agents and robots instead of only making clips.

Genie 4HN front page

Coding-agent control surfaces

Rising · 9 items from 3 sources

Product Hunt again fills with ways to run and watch several coding agents: ManyAgents, AI Bridge, VibeiDE, tabsh and WhoIsCooking each take a slice. PulseDev lets you click a UI bug on localhost and hand it to an agent, and YC's Agent Relay lets a team's agents message each other. On GitHub, openbot and JetBrains' Thinkrail keep trending.

Why now. Teams now keep several long-running agents busy at once, so directing, queuing and reviewing them is daily work that needs its own tools.

PulseDevProduct Hunt

More on this

On-device AI

Rising · 8 items from 5 sources

Liquid AI's LFM3 family is on the Hacker News front page and its 1.7 billion parameter model trends on Hugging Face. Survival LLM packages an assistant and an emergency library for your own computer, Off Grid AI and NINX pitch private AI on phone and laptop, and local-dream runs Stable Diffusion on Android using the Snapdragon NPU.

Why now. Small open models now handle everyday tasks on a phone or laptop, so privacy, offline use and per-token cost become reasons to stay local.

LiquidAI/LFM3-1.7BHF trending models

Anti-slop design skills

Rising · 5 items from 1 sources

Design taste packaged as agent skills is still the fastest-growing corner of GitHub. Hallmark added about 2,800 stars today, more than any other repo, Impeccable added about 1,400, and a new ai-design-skills collection is near 10 thousand. Broader skill frameworks such as Superpowers and ECC sit near 300 thousand stars each.

Why now. Skills became a shared format across Claude Code, Cursor and Codex, so one file of rules reaches every agent a team uses.

Nutlope/hallmarkGitHub Trending

More on this

Verifying agent work

Rising · 4 items from 4 sources

Checking what agents produce is turning into products. Revix AI reviews code against a repo's own rules, a Dev.to plugin stops agents from saying done until tests pass, and a Show HN plugin argues for writing specifications first. Rystic gives agents local copies of Slack, Stripe and Google Drive to practise on.

Why now. Agents now write large changes and take real actions, so the bottleneck has moved from producing work to trusting it.

Revix AIProduct Hunt

Agent-ready Android

Rising · 7 items from 5 sources

Android keeps adding the pieces assistants need. A teardown shows a Gemini connector that can start eSIM setup, AppFunctions is in beta with explainers in Android Weekly and ProAndroidDev, and A2UI Compose for agent-generated UI keeps moving. A new Hugging Face benchmark, iPrefBench, tests personalized phone-use agents.

Why now. Assistants on the phone can now plan multi-step tasks, so apps and system settings need typed entry points for them.

Hate setting up eSIMs? Gemini might soon do it for you (APK teardown)Android Authority

Fading: Agent supply-chain safety (last seen 2026-10-27) · Agents as security auditors (last seen 2026-10-26) · Open voice models (last seen 2026-10-26) · Agent sandboxes (last seen 2026-10-25) · Agent memory layers (last seen 2026-10-21) · Agent harnesses (last seen 2026-10-21) · Open-weight frontier models (last seen 2026-10-21) · AI-found vulnerabilities (last seen 2026-10-21)

Product notes

16Who 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

Genie 4

discussionHN front page792 pts · 484 comments

What
Genie 4, which Google DeepMind's blog calls a new generation of world simulation.
For whom
Researchers training agents and robots, and eventually game and media makers who want explorable scenes from a prompt.
Problem
Training agents in the real world or in hand-built simulators is slow and expensive; teams cope with game engines and limited datasets.
Why now
Video generation now keeps scenes consistent over long interactive runs, which earlier versions could not.
Wedge
A research preview from a well-known series, judged by public demos rather than a product.
Money
Research for now; a path to Google's AI platform and agent training.
Like
Genie 3 · Hy-Nomas · NVIDIA Cosmos · World Labs

LessonWhen a capability is still research, name the job it will do later so people can picture the product.

LiquidAI/LFM3-1.7B

modelHF trending modelstrending 212

What
LFM3-1.7B, a small model from Liquid AI's new LFM3 family, which the company introduces as every modality for specialized intelligence.
For whom
Developers who need a model that fits on a phone, laptop or edge device for one focused task.
Problem
Large hosted models are costly and need a network; small models often handle only text.
Why now
Small models have become good enough for narrow tasks, and devices now ship NPUs to run them.
Wedge
Many input types in a very small model, aimed at specialized uses rather than general chat.
Money
Open weights plus licensing and custom models for companies.
Like
Gemma · Phi · Qwen small models · Gemini Nano

LessonA small model wins by being specific about where it runs and what it is for.

Nutlope/hallmark

repoGitHub Trending★ 2,830 today

What
Hallmark, an anti-AI-slop design skill for Claude Code, Cursor and Codex, at about 39 thousand stars after adding about 2,800 today.
For whom
Developers using coding agents who want generated screens to avoid the recognizable AI look.
Problem
Generated interfaces share tells such as gradients, centered hero cards and filler copy that users now spot.
Why now
Skills became a standard way to extend several coding agents, so one file reaches three tools.
Wedge
Named against a widely felt problem, slop, rather than presented as a general style guide.
Money
Open source, no model yet.
Like
Impeccable · ai-design-skills · shadcn/ui

LessonNaming a pain users already have a word for travels further than describing the feature.

PulseDev

productProduct Hunt#13 on Product Hunt

What
PulseDev lets you click any UI bug on localhost and have your AI agent fix it.
For whom
Front-end developers who already use a coding agent and spend time describing visual bugs in words.
Problem
Telling an agent which element is wrong means screenshots, selectors and long prompts.
Why now
Agents can now edit a codebase on their own, so the slow part is pointing them at the right spot.
Wedge
A single click on the running page as the bug report.
Money
Not stated.
Like
Cursor · Stagewise · Vercel v0

LessonTurning the place where the user notices a problem into the input removes the hardest step.

Paddock

productProduct Hunt#14 on Product Hunt

What
Paddock provides computers and sandboxes for AI agents.
For whom
Teams running agents that need a full machine or isolated environment to work in.
Problem
Letting an agent run code on a laptop or shared server risks damage; teams cope with containers they manage themselves.
Why now
Agents now run long jobs with shell and browser access, which needs disposable, isolated machines.
Wedge
Compute sold to agents as the user rather than to developers.
Money
Likely usage-based compute pricing.
Like
E2B · Daytona · Modal · Rystic

LessonWhen a new kind of user appears, even plain infrastructure can be repackaged for it.

Revix AI

productProduct Hunt#6 on Product Hunt

What
Revix AI, a code reviewer that learns a repo's own rules.
For whom
Teams whose conventions live in reviewers' heads and old pull request comments.
Problem
Generic AI reviewers flag style issues nobody cares about and miss house rules; teams cope with long review checklists.
Why now
Agent-written pull requests multiplied review load, and generic review bots proved noisy.
Wedge
Learning a specific repo's rules rather than applying a universal list.
Money
Likely per-seat or per-repo SaaS.
Like
CodeRabbit · Greptile · Graphite Reviewer

LessonIn a crowded category, personalization to the customer's own data is a clear difference to sell.

Off Grid AI

productProduct Hunt#7 on Product Hunt

What
Off Grid AI, private AI that runs on your phone and computer.
For whom
People who want an assistant without sending chats to a cloud service.
Problem
Hosted assistants need your data and a connection; people cope by sharing less.
Why now
Small open models now run usefully on recent phones and laptops.
Wedge
Privacy and offline use, across both phone and desktop.
Money
Not stated.
Like
NINX · Jan · LM Studio · PocketPal

LessonPrivacy is a wedge only if the product also works when the network does not.

ContextGem 1.0

productProduct Hunt#3 on Product Hunt

What
ContextGem 1.0, which combines decision models with LLMs to extract data from documents.
For whom
Developers building document pipelines such as contract or invoice extraction.
Problem
Pure LLM extraction is hard to audit and repeat; teams cope with long prompts and manual checks.
Why now
LLM extraction is good enough to try in production, which exposes the need for predictable, explainable steps.
Wedge
Pairing deterministic decision models with an LLM instead of trusting one prompt.
Money
Not stated; earlier versions were open source.
Like
LlamaParse · Unstructured · Instructor

LessonPutting a predictable layer around a model is a product in itself when buyers need audits.

OpenRadView

productProduct Hunt#5 on Product Hunt

What
OpenRadView, a free, open-source viewer for your own CT, MRI and X-ray images.
For whom
Patients who receive scans on a disc or download and cannot open them easily.
Problem
Medical images come in formats ordinary computers cannot open; people rely on bundled viewers that often fail.
Why now
Patients increasingly get direct access to their records and want to look at them.
Wedge
Free and open source, aimed at patients rather than hospitals.
Money
Open source, no model yet.
Like
Horos · Weasis · OHIF Viewer

LessonProfessional tools often ignore the end user, and serving that user alone can be enough.

Agent Relay - Let your teams' coding agents message each other

productYC Launches17 votes

What
Agent Relay gives Claude Code, Codex, Cursor, Conductor and other coding agents on a team real-time messaging with each other.
For whom
Engineering teams where each person runs their own agents on a shared codebase.
Problem
Agents working in parallel do not know what others are changing; people relay context by hand in chat.
Why now
Several agents per developer is now common, and they collide on the same files.
Wedge
Vendor-neutral messaging across agents from different companies.
Money
Likely per-team SaaS.
Like
Slack · MCP · ManyAgents

LessonWhen your users' tools come from rival vendors, the neutral layer between them is open ground.

Sentient OS: On-device AI that knows you and takes over your work

productYC Launches3 votesHype flag

What
Sentient OS, on-device AI that learns about your life overnight, takes on tasks with its own cursor and drafts replies in your voice.
For whom
Busy professionals who want an assistant that acts across their apps.
Problem
Assistants need context pasted in and cannot act inside desktop apps; people cope by doing the work themselves.
Why now
Computer-use models and small local models made an on-device agent with its own cursor possible.
Wedge
Overnight learning on the device, so data stays local.
Money
Not stated; likely subscription.
Like
Claude Cowork · Rewind · Apple Intelligence

Claims to know you and take over your work are broad, and the launch shows few votes and no evidence of accuracy.

LessonThe bigger the promise about knowing the user, the more a narrow first task is needed to earn trust.

Survival LLM: AI assistant and emergency library that runs on your own computer

discussionHN front page235 pts · 116 comments

What
Survival LLM, an AI assistant and emergency library that runs on your own computer.
For whom
Preppers, travellers and people in places with unreliable internet.
Problem
Help and reference material vanish when the network does; people keep printed guides or offline wikis.
Why now
A capable model and a reference library now fit on an ordinary laptop.
Wedge
Framed for emergencies, a setting where offline is a requirement, not a preference.
Money
Not stated.
Like
Kiwix · Internet-in-a-Box · Jan

LessonPick the situation where your constraint is a must-have, and the product explains itself.

Nuvio 5.0

productProduct Hunt#1 on Product Hunt

What
Nuvio 5.0 lets you hand your financial admin to your AI agent; it was the top Product Hunt launch.
For whom
Freelancers and small business owners who dread bookkeeping and invoices.
Problem
Financial admin is repetitive and easy to put off; people cope with spreadsheets or an accountant.
Why now
People now run personal agents, so a product can be built for the agent to operate rather than the person.
Wedge
Positioned as a tool your existing agent uses, not another app to learn.
Money
Not stated; likely subscription.
Like
QuickBooks · Pilot · Xero

LessonDesigning for the user's agent as the operator is a new way to enter an old category.

xororz/local-dream

repoGitHub Trending · KotlinAndroid★ 10 today

What
local-dream runs Stable Diffusion on Android devices with Snapdragon NPU acceleration, with CPU and GPU support as well.
For whom
Android users and developers who want image generation on the phone without a server.
Problem
Image generation usually needs a cloud GPU or a desktop card; on phones it was too slow to use.
Why now
Recent Snapdragon NPUs are fast enough for diffusion models, and Qualcomm's toolchains made them reachable.
Wedge
NPU acceleration on a specific chip family, where the speed difference is largest.
Money
Open source, no model yet.
Like
Draw Things · MLC LLM · Qualcomm AI Hub

LessonTargeting one hardware path well beats supporting every device poorly when speed is the selling point.

jieapi/AiCode

repoGitHub Trending · KotlinAndroid★ 3 today

What
AiCode, an AI coding agent that runs on Android with a built-in Linux terminal and code editor, extended through MCP.
For whom
Developers who want to code or fix small things from a phone or tablet.
Problem
Coding on Android means juggling Termux, an editor and a separate AI chat.
Why now
Coding agents and MCP became standard, so an agent can be packaged with its own tools in one app.
Wedge
Terminal, editor and agent bundled in a single Android app.
Money
Open source, no model yet.
Like
Termux · Claude Code · Replit mobile

LessonBundling the tools an agent needs makes a new platform viable before the big vendors arrive.

Introducing Crayfish: An Image Cropper for Compose Multiplatform

articleDove Letter (skydoves)Android

What
Crayfish, an image cropper for Compose Multiplatform with a one-line crop dialog, a Cropper composable, aspect ratios, gestures, shapes and Coil support.
For whom
Android and Kotlin Multiplatform teams that need avatar or photo cropping on several platforms.
Problem
Most Android croppers are View-based or Android-only, so multiplatform apps write their own.
Why now
Compose Multiplatform is stable on iOS, so shared UI libraries for common tasks now have a market.
Wedge
A one-line dialog for the common case, with a composable for custom layouts.
Money
Open source, no model yet.
Like
uCrop · Android-Image-Cropper · Krop

LessonOffer a one-line default and an escape hatch, and you serve both the hurried and the picky.

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

11
  • elayadesign/ai-design-skillsnew

    A collection of AI design skills for coding agents, near 10 thousand stars after about 1,400 today.

    ★ 1,395 today

  • Nutlope/hallmarkcard

    Anti-AI-slop design skill for Claude Code, Cursor, and Codex.

    ★ 2,830 today

  • nightly-labs/openbot

    openbot, a local-first workspace for persistent agent teammates across Codex, Claude and Grok.

    ★ 777 today

  • obra/superpowersnew

    Superpowers, an agentic skills framework and development method, now past 300 thousand stars.

    ★ 617 today

  • pbakaus/impeccable

    Impeccable, a design vocabulary for coding agents, passes 90 thousand stars as design taste becomes an installable file.

    ★ 1,449 today

  • makecindy/cindy

    Cindy, an open-source agent pitched as working out of the box, keeps trending.

    ★ 1,465 today

  • gitroomhq/postiz-app

    Postiz, an open-source social media scheduler that now pitches itself as agent-driven.

    ★ 951 today

  • affaan-m/ECC

    ECC bundles skills, memory and security checks for agent harnesses, near 290 thousand stars.

    ★ 1,165 today

Show 3 more
  • HKUDS/DeepTutornew

    DeepTutor, an open-source personalized tutoring system, added about 1,600 stars today.

    ★ 1,598 today

  • JetBrains/thinkrail

    JetBrains' Thinkrail, a lightweight agentic IDE, keeps climbing with about 1,200 stars today.

    ★ 1,204 today

  • togg53192-cmd/jailbreaksnew

    LIST OF ALL MY JAILBREAKS

    ★ 902 today

Product Hunt

50
  • Nuvio 5.0newcard

    Hand your financial admin to your AI agent

    #1 on Product Hunt

  • GPT-6.1 Terranew

    GPT-6.1 Terra, OpenAI's cheaper computer-use model, is still near the top of Product Hunt.

    #2 on Product Hunt

  • ContextGem 1.0newcard

    Decision models join LLMs to extract data from documents

    #3 on Product Hunt

  • Coacharenew

    A personal coaching app with a generic pitch and no stated method.

    #4 on Product Hunt

  • OpenRadViewnewcard

    Free, open-source viewer for your own CT, MRI and X-rays

    #5 on Product Hunt

  • Revix AInewcard

    AI code review that learns your repo's own rules

    #6 on Product Hunt

  • Off Grid AInewcard

    Private AI on your phone and computer.

    #7 on Product Hunt

  • NuPlaynew

    A platform for running a company's AI transformation, broad and light on specifics.

    #8 on Product Hunt

Show 42 more
  • SavorApp_AInew

    Captures notes and intentions and turns them into next steps.

    #9 on Product Hunt

  • Spacesheepnew

    Spacesheep's AI-made documents for teams, also a YC launch today.

    #10 on Product Hunt

  • TheySaid 4.0new

    Simulates how users will react to a design right inside the build tool.

    #11 on Product Hunt

  • Orbitnew

    Makes you earn each app opening to curb doomscrolling, digital wellbeing as a game.

    #12 on Product Hunt

  • PulseDevnewcard

    Click any UI bug on localhost and your AI agent fixes it

    #13 on Product Hunt

  • Paddocknewcard

    Computers and sandboxes for your agents

    #14 on Product Hunt

  • fswitchernew

    A replacement for Command-Tab app switching on the Mac.

    #15 on Product Hunt

  • Overmindnew

    An open-source platform for continuous learning in AI systems.

    #16 on Product Hunt

  • NINXnew

    Private on-device AI with connected tools and creative workflows.

    #17 on Product Hunt

  • Notranew

    A tool for ranking in AI search answers, the new SEO.

    #18 on Product Hunt

  • Arcade Controller Pronew

    A fight-game controller with buttons instead of a stick and its own screen.

    #19 on Product Hunt

  • iPad mini 8new

    Apple's iPad mini 8 with OLED, A20 Pro and IP67.

    #20 on Product Hunt

  • AI Bridge

    Keeps every AI coding agent on your Mac organized in one place.

    #21 on Product Hunt

  • Knock

    Trigger Mac shortcuts by tapping the laptop body.

    #22 on Product Hunt

  • Loqui

    Turns your voice into ready-to-send text in any Mac app.

    #23 on Product Hunt

  • WhoIsCooking

    Shows which AI agent is heating your Mac and lets you cool it down.

    #24 on Product Hunt

  • omerta

    Catches .env secrets before they leak into git.

    #25 on Product Hunt

  • desigma

    A free Figma alternative written in Rust.

    #26 on Product Hunt

  • ManyAgents

    Runs a team of AI coding agents from one Mac window.

    #27 on Product Hunt

  • Mozy

    An app blocker for Android and iPhone.

    #28 on Product Hunt

  • VibeiDE

    Queue agent tasks in parallel and review each result.

    #29 on Product Hunt

  • Standfast

    Tells you whether your Mac's self-hosted GitHub Actions runners will pick up jobs.

    #30 on Product Hunt

  • AVP

    An open-source Git alternative for AI coding agents, a large claim with little detail.

    #31 on Product Hunt

  • AllModels

    Explore every abliterated model

    #32 on Product Hunt

  • 2G

    A sampler for early-2000s neural synthesis sounds.

    #33 on Product Hunt

  • MiniMax M3.1 Flash

    MiniMax M3.1 Flash, claimed three times faster and 70 percent cheaper for agents.

    #34 on Product Hunt

  • Flipcorder

    A 1991 camcorder look for the iPhone, one tape a day.

    #35 on Product Hunt

  • Strawberry 1.0

    An AI browser for teams claiming over 100 saved hours a week, with no evidence in the listing.

    #36 on Product Hunt

  • PDF Accessibility Checker by WebYes

    A free PDF accessibility checker with fix guidance.

    #37 on Product Hunt

  • Kiwi

    A proactive assistant that texts you before you ask.

    #38 on Product Hunt

  • Helio

    Shows what your rooftop solar produced today.

    #39 on Product Hunt

  • tabsh

    An IDE in a browser tab for working with agents.

    #40 on Product Hunt

  • N8N Chat UI

    No-code branded chat widgets for n8n workflows.

    #41 on Product Hunt

  • CloudKnife

    Cloud cost cutting sold as a product.

    #42 on Product Hunt

  • Jobmate AI

    An AI copilot for the whole job search.

    #43 on Product Hunt

  • SurgePix

    AI visuals where every layer stays editable and agents can call the tool.

    #44 on Product Hunt

  • Moli by Lexmount

    A headless browser built for agent speed.

    #45 on Product Hunt

  • Clips by Plastic

    Generates lyric video ads timed to the beat for musicians.

    #46 on Product Hunt

  • Badgey

    Turns an iPhone into a name badge that faces the room.

    #47 on Product Hunt

  • Llama Windows

    One local AI server that every Windows app can call.

    #48 on Product Hunt

  • Openstatus Incident Management

    Incident management sized for small teams.

    #49 on Product Hunt

  • Yourcast 2.0

    AI podcast summaries with tap-to-jump timestamps inside Telegram.

    #50 on Product Hunt

Show HN

3

HN front page

30
Show 22 more

YC Launches

20
Show 12 more

HF trending models

18
Show 10 more

HF trending spaces

16
Show 8 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

388 items from 77 sources, 11 of them seen in several sources. 3 sources failed to fetch.

Source health

77Which sources were fetched today, and how many items each returned.
SourceSectionItemsStatus
GitHub TrendingMain11Fetched
Product HuntMain50Fetched
Show HNMain3Fetched
HN front pageMain30Fetched
YC LaunchesMain20Fetched
HF trending modelsMain20Fetched
HF trending spacesMain20Fetched
HF daily papersMain20Fetched
LobstersMain25Fetched
Dev.to topMain20Fetched
TechmemeMain15Fetched
AndroidX releasesAndroid · Official & releases46Fetched
Android Developers BlogAndroid · Official & releases2Fetched
Android Studio releasesAndroid · Official & releases4Fetched
Kotlin blogAndroid · Official & releases3Fetched
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 & releases0Fetched
AndroidX merged API changesAndroid · AOSP / AndroidX insider27Fetched
Dove Letter (skydoves)Android · Experts5Fetched
skydoves on MediumAndroid · Experts0Fetched
Jake WhartonAndroid · Experts0Fetched
Zac SweersAndroid · Experts0Fetched
Chris BanesAndroid · Experts3Fetched
Ryan HarterAndroid · Experts0Fetched
Talking KotlinAndroid · Experts0Fetched
Joe BirchAndroid · Experts2Fetched
CommonsWare (Mark Murphy)Android · Experts3Fetched
TechYourChance (Vasiliy Zukanov)Android · Experts1Fetched
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 releases1Fetched
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 releases0Fetched
kotlinx.coroutines releasesAndroid · Library releases0Fetched
kotlinx.serialization releasesAndroid · Library releases1Fetched
Ktor releasesAndroid · Library releases0Fetched
Apollo Kotlin releasesAndroid · Library releases0Fetched
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 releases0Fetched
MockK releasesAndroid · Library releases0Fetched
Lottie releasesAndroid · Library releases0Fetched
Authority Insights (leaks & teardowns)Android · Leaks & ecosystem7Fetched
9to5Google · AndroidAndroid · Leaks & ecosystem8Fetched
Android AuthorityAndroid · Leaks & ecosystem6Fetched
Android FaithfulAndroid · Leaks & ecosystem2Fetched
Android WeeklyAndroid · Community15Fetched
GitHub Trending · KotlinAndroid · Community11Fetched
Lobsters · androidAndroid · Community0Fetched
HN · AndroidAndroid · Community12Fetched
ProAndroidDevAndroid · Community5Fetched
Kotlin WeeklyAndroid · Community1Fetched
Philipp Lackner (YouTube)Android · Community0Failed
1 of 5

Grab cut ANRs by a third with R8 and app profiles

If your app still uses old ProGuard rules or has no Baseline Profile, this is a measured case for fixing both.

The Android Developers Blog describes how Grab, the Southeast Asian superapp, tuned its build. It first switched on R8 optimization, then moved to R8 full mode, and added custom Baseline and Startup Profiles for its startup paths. Overall, freezes known as ANRs fell about 32 percent, cold start improved about 35 percent, and the app shrank about 30 percent.

Android Developers Blog · OfficialRead the source

1 of 4

DeepMind's Genie 4 pushes world simulation forward

Generated worlds are becoming a cheap place to train and test agents and robots before they touch the real one.

Google DeepMind announced Genie 4, which its blog calls a new generation of world simulation. World models generate an environment you can move through and act in, frame by frame, rather than a fixed video. It was one of the biggest Hacker News stories of the day, with about 790 points and nearly 500 comments. Several top Hugging Face papers also studied world models for robots.

HN front page · Worth a debateRead the sourceDiscussion