Android Feed

← The 5 stories that mattered

Everything collected ·

All 334 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. Gemini 4 Argon arrives for Android developers

    Agents can now drive your local Android build and device tools through an official plugin, not only suggest code.

    Google brought its newest model, Gemini 4 Argon, to Android development. You can use it in Android Studio Canary by signing into Antigravity, Google's agent app, or through the paid Gemini API. Google also shipped an Android CLI plugin that connects any terminal agent to your local Android tools. The Antigravity route needs a Google AI Ultra plan, and the post gives no benchmark numbers.

    Android Developers Blog · OfficialRead the source

  2. Grab cuts freezes and slow starts with R8 and app profiles

    If your app still ships without full R8 optimization or baseline profiles, this is a measured case for turning them on.

    Grab, a large Southeast Asian super app, says it reduced app-not-responding errors by 32 percent and made startup 35 percent faster. It did this with R8, the tool that shrinks and optimizes Android code at build time, and with profile-guided optimization, which tells the system which code to prepare ahead of time. The post on the Android Developers Blog explains the steps.

    Android Developers Blog · OfficialRead the source

  3. Late Android source releases push developers to disassembly

    If you write native code, you may not be able to check device behavior against public sources for a while.

    A developer found a bug in Android 17's C library on Pixel phones, but the public source code was older than the builds on devices. To report it, they took apart the compiled library. Google's engineer confirmed the bug and promised tests, a fix and a backport. The thread, popular on Hacker News, also questions where newer Android 17 sources are.

    HN · Android · CommunityRead the sourceDiscussion

  4. Google's Android skills for coding agents top the Kotlin charts

    Adding these skills is a cheap way to make your coding agent write current Android code instead of outdated patterns.

    Google's android skills repository is today's top trending Kotlin project on GitHub. It holds instruction files that teach coding agents how to do Android tasks the way the official docs recommend, such as analyzing R8 output, using CameraX, casting with Media3 and profiling. It installs through the Android command line tool, or as a plugin for Claude Code and Codex.

    GitHub Trending · Kotlin · CommunityRead the sourceProduct notes

  5. Haze 2.0 brings glass effects to Compose

    If your app uses frosted-glass or blurred headers, check whether built-in Compose blur now covers you before adding a dependency.

    Chris Banes released Haze 2.0, a popular blur library for Compose Multiplatform. It adds Glass, a new effects interface, performance improvements, and experimental support for Android's native backdrop blur. In a second post he explains how Compose's new built-in progressive blur relates to Haze, and when you still need the library.

    Chris Banes · ExpertRead the source

Also worth knowing

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

  • kotlinx.serialization 1.12.0

    kotlinx.serialization 1.12.0 is stable, built on Kotlin 2.4.10, with a stable Uuid serializer and cleaned-up ProGuard rules.

    Any Kotlin app using kotlinx.serialization; plan the upgrade together with Kotlin 2.4.10.

    Adopt nowOfficial

  • Start Your Next KMP App With Kotlin Toolchain 0.13

    JetBrains now recommends Kotlin Toolchain 0.13, a declarative build entry point, for most new Kotlin Multiplatform projects.

    Teams starting new KMP apps; existing Gradle projects need not move yet.

    WatchOfficial

Go deeper: every Android item, by lane

Official & releases

20Gemini 4 Argon and an Android CLI plugin for agents land in Android Studio Canary, Grab's R8 and profile results make the case to finish optimizing your build, and Compose Multiplatform 1.12.2 is a safe fix release.

AndroidX library releases

Stable · safe to ship · 2

Beta · API frozen, try it before it ships · 1

Alpha · experimental, APIs may change · 2
Show 7 more

AOSP / AndroidX insider

11Gerrit shows Remote Compose gaining animations and layers, AppFunctions dropping its legacy runtime, and Compose preview tooling for agent-generated UI.
Show 3 more
  • AppFunctions: legacy runtime removednew

    Remove legacy AppFunction runtime classes and aggregated inventory

    AppFunctions, the API that lets agents call into apps, is consolidating on its new design; old integration paths are going away.

  • AppFunctions: old aggregation mode removednew

    Remove legacy AppFunction aggregation mode

    Another cleanup of AppFunctions, suggesting the API surface is settling before a stable release.

  • Compose snapshots: smarter null checks in readOnlynew

    Add callsInPlace contract to Snapshot.readOnly

    A Kotlin contract on Snapshot.readOnly lets the compiler understand code inside it, a small quality fix for state-heavy code.

Experts

15Read Chris Banes on Haze 2.0 and Compose's new progressive blur, and skydoves' two-part series on how the Compose compiler and runtime work.
Show 7 more

Library releases

4kotlinx.serialization 1.12.0 and SQLDelight 2.4.1 are stable and worth updating; Decompose 3.6.0 is still beta.
  • Compose Unstyled 2.10.1

    Patch Changes 9658157 : Support dismissOnClickOutside for DrawerPresentation.InPlace

  • SQLDelight 2.4.1

    Added [Documentation] Additional Android SQLite transaction documentation ( #6381 by @griffio ) Changed [Gradle Plugin] Keep parsed .sq files in memory for the whole code generation task, so they are not parsed again after garbage collection. This can make code generation faster…

  • kotlinx.serialization 1.12.0new

    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…

Leaks & ecosystem

18Leaks point to work-profile apps in Android Auto, Gemini replacing Ok Google on XR glasses, and new Google Wallet pass features.
Show 10 more

Community

41Hacker News is debating late AOSP source releases and Mac.apk, which runs Android apps natively on macOS.
Show 33 more
Start the Product & AI read4 stories, ranked by what you’d regret missing · 2 min read

  1. A local AI job search kit is GitHub's fastest riser

    It shows that a well-packaged set of agent workflows can spread like a product without a server or a company behind it.

    An open-source job search framework built on Claude Code gained almost four thousand stars in a day and now has about fifty eight thousand. It evaluates job postings, tailors your CV, writes cover letters and helps you prepare for interviews. Everything runs on your own machine, and the author encourages people to fork it and adapt it.

    GitHub Trending · Fastest risingRead the sourceProduct notes

  2. Developers debate shipping AI-written code they don't understand

    Review and understanding, not code generation, are becoming the limit on how fast teams can safely ship.

    An Ask Hacker News thread drew more than four hundred comments on how to handle code written by AI that the author does not fully understand. The same day, Lobsters members argued over rules for posts tagged as vibe coding, and a Dev.to post warned that teams are cutting the review steps that catch risky changes.

    HN front page · Worth a debateRead the source

  3. Tools for managing many coding agents at once multiply

    With agents commoditized, the place where people supervise them is becoming the product people choose.

    Orca, a workspace for running a fleet of coding agents with your own subscriptions, gained about sixteen hundred GitHub stars today. Product Hunt also featured several tools in the same space: an infinite canvas for agents, a buddy-list view, a walkie-talkie interface, and an app that tells you when to push your agents harder or ease off.

    GitHub Trending · Across sourcesRead the sourceProduct notes

  4. Inception's Mercury 3 bets on diffusion for faster text

    If the speed holds up, latency-sensitive features like autocomplete and voice could use stronger models than before.

    Inception launched Mercury 3, a language model that writes text by diffusion, refining a whole answer at once instead of one word at a time. The company claims quality comparable to small mainstream models while running four to eight times faster, at about one thousand tokens per second. The claims come from the company and have not been independently tested here.

    Techmeme · New ideaRead the source

Go deeper: themes, product notes, every source

Themes across sources

6Coding-agent control surfaces · Open-weight frontier models · Agent harnesses · Agent memory layers · Agent supply-chain safety · Agent-ready Android

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

Coding-agent control surfaces

Rising · 8 items from 3 sources

The fight has moved from the agent to the room you run agents in. Orca, a desktop and mobile home for a fleet of coding agents, gained about 1,600 GitHub stars today, while Product Hunt launched Starkomand's infinite canvas, buddylist's chat-style agent list, a walkie-talkie interface, Lancetta's pacing advice and Beam.win for driving agents on another Mac.

Why now. Developers now run several coding agents at once on their own subscriptions, so watching, steering and switching between them is the new bottleneck.

stablyai/orcaGitHub Trending

Open-weight frontier models

Rising · 6 items from 3 sources

Very large open-weight models keep arriving. Product Hunt featured Beam at 501 billion parameters and Step 5 Preview at 600 billion, both pitched at agent work, while DeepSeek V4.1 Flash tops Hugging Face trending and MiniMax M3.1 Flash drew over 400 points on Hacker News.

Why now. Mixture-of-experts designs keep the active size small, so labs can release huge models that are still practical to host and resell.

deepseek-ai/DeepSeek-V4.1-Flash-1017HF trending models

More on this

Agent harnesses

Rising · 6 items from 3 sources

The code wrapped around a model is being treated as a product and a research object. Ruflo and CLI-Anything trend on GitHub, a YC launch proposes UHP as an open standard for running whole harnesses as a backend, and two Hugging Face papers optimize the harness itself alongside the model.

Why now. Models are good enough that tool wiring, memory and retries now decide most of an agent's results, so teams are standardizing and tuning that layer.

HKUDS/CLI-AnythingGitHub Trending

Agent memory layers

Steady · 5 items from 2 sources

Memory is splitting off from the agent. ai-memory on GitHub keeps long-term memory that moves between coding CLIs from different vendors, Gravity Notes shares one note list between you and your agents, and Amazon's Bee devices and Wally pitch personal memory that learns how you write and live.

Why now. People switch between agents and vendors often, so memory tied to one tool loses value; a portable layer keeps context when the tool changes.

akitaonrails/ai-memoryGitHub Trending

More on this

Agent supply-chain safety

New · 5 items from 3 sources

What goes into and out of agents is becoming a security topic. shim-cli shows and masks what a coding agent sends, a Dev.to post runs a logging proxy in front of an LLM client, RepoPilot scores repos after thousands of fake ones posed as skills and MCP servers, and Hacker News discussed abliterated models that can be turned against users and hidden LLM instructions on WhatsApp's help pages.

Why now. Agents now read web pages, install skills and send code to remote models on their own, so hidden instructions and leaked data have a direct path into real work.

shim-cliProduct Hunt

Agent-ready Android

Rising · 5 items from 4 sources

Google keeps packaging Android knowledge for agents. Gemini 4 Argon arrived in Android Studio Canary with an Android CLI plugin for any terminal agent, the android/skills repo of agent instructions is the top Kotlin repo on GitHub, a Developer Knowledge API and MCP server serves Android docs to assistants, and Gerrit shows the AppFunctions runtime being cleaned up.

Why now. Agents now run builds and devices end to end, and Google wants them to follow current Android guidance rather than stale training data.

Bringing our latest frontier model to Android DevelopersAndroid Developers Blog

Fading: On-device AI (last seen 2026-10-19) · Verifying agent work (last seen 2026-10-19) · Agent sandboxes (last seen 2026-10-17) · Small decision models (last seen 2026-10-16) · AI-driven reverse engineering (last seen 2026-10-16) · MCP-native apps (last seen 2026-10-13)

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

MadsLorentzen/ai-job-search

repoGitHub Trending★ 3,856 today

What
An open-source job search framework built on Claude Code that evaluates postings, tailors CVs, writes cover letters and preps interviews, all on your own machine.
For whom
Job seekers comfortable with a terminal who apply to many roles and want each application tailored.
Problem
Tailoring a CV and letter for every posting takes hours, so people send generic applications or pay for hosted tools that keep their data.
Why now
Coding agents can now run long, multi-step workflows over local files, so a set of prompts and scripts can act like a full product.
Wedge
Fork it and run it locally: your CV and search history never leave your machine.
Money
Open source, no model yet; it depends on the user's own Claude subscription.
Like
Teal · Huntr · Jobscan

LessonA repo of agent workflows can be a product: the agent provides the engine, you ship the process and the domain knowledge.

stablyai/orca

repoGitHub Trending★ 1,590 today

What
Orca, a workspace for running a fleet of coding agents such as Grok Bot and Claude Code from one place, on desktop and mobile.
For whom
Developers who run several coding agents in parallel and pay for their own agent subscriptions.
Problem
Each agent lives in its own terminal or app, so juggling many means many windows and lost track of what each is doing.
Why now
Running agents side by side on long tasks became normal this year, which makes supervision the bottleneck instead of generation.
Wedge
Bring any agent with your own subscription, so there is no new model bill to justify.
Money
Open source repo; no paid plan is described in the listing.
Like
Conductor · Claude Code desktop · Starkomand · tmux

LessonWhen the core engine is a commodity you can't own, own the place where people manage it.

headroomlabs-ai/headroom

repoGitHub Trending★ 310 todayHype flag

What
Headroom, a library that compresses tool outputs, logs, files and retrieved chunks before they reach a language model.
For whom
Teams running coding agents or retrieval pipelines that pay per token and hit context limits.
Problem
Agents stuff raw logs and JSON into the prompt, wasting tokens and money; teams cope by hand-trimming or truncating.
Why now
Agent runs now read huge tool outputs in loops, so token spend scales with tool noise, not with the task.
Wedge
A drop-in layer with a concrete claim: about 20 percent fewer tokens for coding agents and 60 to 95 percent fewer for JSON, with the same answers.
Money
Open-source library; a hosted or enterprise tier is plausible but not stated.
Like
LLMLingua · prompt caching

Same answers is a strong claim; the listing gives savings figures but not how answer quality was measured.

LessonPick the costliest waste in a new workflow and sell a measurable percentage cut on it.

akitaonrails/ai-memory

repoGitHub Trending★ 666 today

What
ai-memory, long-term memory for coding agent CLIs that also helps hand work off between agents from different vendors.
For whom
Developers who switch between Claude Code, Codex, Gemini and similar CLIs on the same project.
Problem
Each agent forgets past sessions and knows nothing of what another agent did, so people keep notes files and repeat context by hand.
Why now
Developers now use several vendor CLIs at once, so memory tied to one vendor no longer covers their work.
Wedge
Cross-vendor handoff, which no single agent vendor has a reason to build.
Money
Open source, no model yet.
Like
Mem0 · Letta · CLAUDE.md files

LessonBuild the feature each vendor has a reason not to build: interoperability between rivals.

Beam

productProduct Hunt#3 on Product HuntHype flag

What
Beam, an open model with 501 billion parameters described as highly efficient and built for agent work.
For whom
Teams that want to host or fine-tune a large agent model instead of renting a closed one.
Problem
The best agent models are closed, which limits control over cost, hosting and data.
Why now
Mixture-of-experts training made very large open models affordable to release and run.
Wedge
Open weights plus an efficiency pitch, launched to a product audience rather than only researchers.
Money
Not stated; likely hosted API access alongside free weights.
Like
Step 5 Preview · DeepSeek V4.1 · Muse Spark Open · Qwen3.8

The listing says highly efficient but gives no benchmarks, active size or cost figures.

LessonIn a crowded field, one number like 501B gets attention, but buyers will ask for the comparison you left out.

Point and Yap

productProduct Hunt#5 on Product Hunt

What
Point and Yap lets you point at something on screen and talk to your agent about it, passing the context through MCP.
For whom
People using desktop agents who find typing out screen context slow.
Problem
Describing a bug or UI element to an agent means screenshots and long text; pointing and speaking is faster.
Why now
MCP lets a small tool feed context to any compatible agent, and voice input is now accurate enough to use all day.
Wedge
One gesture plus voice, delivered as an MCP server so it works with agents people already have.
Money
Not stated.
Like
Wispr Flow · Superwhisper · macOS screenshot tools

LessonShipping as a plug-in to a standard like MCP lets a tiny tool reach every agent without picking a winner.

shim-cli

productProduct Hunt#38 on Product Hunt

What
shim-cli shows what your coding agent sends to the model provider and masks what you choose to hide.
For whom
Developers and security teams who worry about secrets or proprietary code leaving through agent prompts.
Problem
Agent traffic is opaque; teams either trust the vendor or block agents entirely.
Why now
Agents read whole repos and environment files on their own, so leaks can happen without anyone pasting anything.
Wedge
Visibility first: a local view of outgoing requests, with masking as the next step.
Money
Not stated; team policy and audit features would be a natural paid tier.
Like
mitmproxy · LLM gateways · secret scanners

LessonShowing people a problem they suspected but couldn't see is a strong first feature for a security tool.

Starkomand

productProduct Hunt#15 on Product Hunt

What
Starkomand runs your AI coding agents on an infinite canvas, so each agent is a tile you can arrange and watch.
For whom
Developers running several agents who think spatially rather than in tabs.
Problem
Terminal tabs hide which agent is waiting for input or stuck.
Why now
Parallel agent work is common enough that layout and monitoring are their own product.
Wedge
A visual canvas, a different shape from the list and tab views other managers use.
Money
Not stated.
Like
Orca · buddylist · tldraw computer

LessonWhen a category fills up fast, a distinct interface metaphor is a cheap way to be remembered.

Show HN: Qwen3-4B with the AI writing tics trained out

productShow HN145 pts · 108 comments

What
A fine-tuned Qwen3-4B model with common AI writing tics trained out, shared as a Show HN.
For whom
Writers and app builders who want small-model text that doesn't read as machine-written.
Problem
Model text repeats stock phrases readers now spot instantly; people cope with long style prompts or manual edits.
Why now
Small models are cheap to fine-tune, and readers' fatigue with AI phrasing is now a known problem.
Wedge
A small open model aimed at one narrow quality, style, rather than general ability.
Money
Hobby release, no model.
Like
humanizer models · style prompts

LessonRemoving one widely felt annoyance can earn more attention than a general capability gain.

Show HN: KiviDB, a faster Redis-compatible memory database and vector cache

productShow HN66 pts · 39 comments

What
KiviDB, a Redis-compatible in-memory database and vector cache written in Rust, built to avoid garbage-collection pauses.
For whom
Backend teams using Redis who also need a vector cache for AI features.
Problem
Teams run Redis plus a separate vector store, adding cost and moving parts.
Why now
Caching embeddings and model answers is now a routine need next to ordinary key-value caching.
Wedge
Redis compatibility, so existing clients work without code changes.
Money
Open source side project, no model yet.
Like
Redis · Valkey · Dragonfly · KeyDB

LessonSpeak an existing protocol and switching costs drop to near zero, which is the easiest wedge for infrastructure.

PlaceCall by VOYGR – The phone is the last API

productYC Launches104 votes

What
PlaceCall by VOYGR, an API that finds the right businesses, calls them in parallel and returns answers, quotes and bookings.
For whom
Developers building agents or apps that need information only available by phone.
Problem
Many local businesses have no API or online booking, so the last step needs a human on the phone.
Why now
Voice agents can now hold short, useful phone calls, so the phone can be wrapped as an API.
Wedge
The pitch that the phone is the last API, sold to developers rather than end users.
Money
Likely usage-based API pricing; not stated.
Like
Bland AI · Vapi · Google Duplex

LessonFind the step in a workflow that still needs a person and wrap it in an API the rest of the stack can call.

Veeza AI (F26) — AI tourist visa agent for people with weak passports

productYC Launches70 votes

What
Veeza AI, a YC company that acts as a tourist visa agent, handling the application once you pick a destination.
For whom
Travelers with passports that need visas for most destinations.
Problem
Visa paperwork is slow and error-prone; people use costly agents or risk rejection doing it alone.
Why now
Agents can now fill long forms and assemble documents reliably enough to sell the outcome, not software.
Wedge
A clear, underserved audience: people with weak passports.
Money
Likely a per-application service fee; not stated.
Like
iVisa · VisaHQ · Atlys

LessonDefining the customer by their pain, here a weak passport, makes the marketing write itself.

Android apps running native on macOS

discussionHN front page, HN · AndroidAndroid284 pts · 153 commentsHype flag

What
Mac.apk, an alpha open-source project that runs Android apps on Apple Silicon Macs as native macOS processes, without an emulator or virtual machine.
For whom
Mac users who want Android-only apps, and Android developers curious about running their app on desktop.
Problem
Running an Android app on a Mac today means a full emulator, which is heavy and feels out of place.
Why now
It targets macOS 26.4 or newer on Apple Silicon, and builds the Android app environment directly on top of macOS.
Wedge
Each app gets its own Dock icon and window, so it feels like a Mac app rather than a phone in a box.
Money
Open source under AGPL-3.0, no model.
Like
Android Emulator · Waydroid · BlueStacks · Windows Subsystem for Android

It is an alpha; how many real apps work, especially ones that need Google Play services, is not shown.

LessonTranslating an environment rather than emulating hardware trades completeness for a much better feel.

android/skills

repoGitHub Trending · KotlinAndroid★ 323 today

What
android/skills, Google's repo of instruction files that teach coding agents Android tasks from official guidance, such as R8 analysis, CameraX, Media3 casting and profiling.
For whom
Android developers who use Claude Code, Codex, Gemini or Antigravity on Android projects.
Problem
Agents write outdated or wrong Android code in areas where models are weak; developers fix it by hand.
Why now
Skill files became a common format across agent tools, so one set of instructions can serve several agents.
Wedge
Installs through the Android CLI or as a plugin for Claude Code and Codex, not only Google's own agent.
Money
Free; it supports Google's Android and Gemini platforms.
Like
Android docs · Cursor rules · Context7

LessonPlatform owners can shape agent output by publishing guidance in the agent's own format, even for rival agents.

jqssun/android-airplay-server

repoGitHub Trending · KotlinAndroid★ 26 today

What
android-airplay-server, described as the first open-source AirPlay receiver for Android, with screen mirroring, video and audio.
For whom
People who want to cast from an iPhone or Mac to an Android TV or tablet.
Problem
AirPlay mostly targets Apple TV and licensed TVs; Android users rely on closed paid apps.
Why now
Not clear from the listing; it is a fresh trending repo in the Kotlin list.
Wedge
Being the first open-source option for a common household need.
Money
Open source, no model.
Like
AirScreen · UxPlay · Apple TV

LessonBridging two ecosystems that refuse to talk is a lasting source of demand.

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

12
  • MadsLorentzen/ai-job-searchcard

    The job search that runs on your machine. AI job application framework built on Claude Code: evaluate postings, tailor CVs, write cover letters, prep interviews. Fork it and own it.

    ★ 3,856 today

  • anomalyco/opencode

    The leading open-source coding agent keeps gaining stars, a sign developers want an agent not tied to one model vendor.

    ★ 1,311 today

  • akitaonrails/ai-memorycard

    Solution for long term memory for agent coding CLIs and to facilitate handoff between different agent vendors

    ★ 666 today

  • Comfy-Org/ComfyUI

    The node-based image generation app keeps trending as new image models ship Comfy workflows on day one.

    ★ 1,011 today

  • ruvnet/ruflo

    A popular agent orchestration framework for multi-agent swarms; big claims, but a clear sign of interest in harnesses.

    ★ 564 today

  • stablyai/orcacard

    Orca is the place for working with a fleet of agents (e.g. Grok Bot, Claude Code). Run any coding agent with your own subscription. Available on desktop, mobile and remote runtime.

    ★ 1,590 today

  • headroomlabs-ai/headroomnewcard

    Compress tool outputs, logs, files, and RAG chunks before they reach the LLM. 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers. Library, proxy, MCP server.

    ★ 310 today

  • withmarbleapp/os-taxonomynew

    A trending taxonomy repo with no description yet; worth a look only if you classify operating systems or apps.

    ★ 284 today

Show 4 more
  • Kruk2/jasna

    An open-source video restoration tool climbing GitHub, part of the steady move of video cleanup to local machines.

    ★ 498 today

  • HKUDS/CLI-Anything

    Wraps ordinary software in command-line interfaces so agents can drive it, betting on CLIs as the agent interface.

    ★ 606 today

  • anthropics/claude-codenew

    Anthropic's terminal coding agent remains one of the most starred agent repos.

    ★ 167 today

  • shiyu-coder/Kronosnew

    A foundation model trained on financial market data, applying the language-model recipe to price series.

    ★ 363 today

Product Hunt

49
  • Lookout (for Slack)new

    Filters Slack for urgent messages and acts on them, aiming at notification overload in busy workspaces.

    #1 on Product Hunt

  • Beamnewcard

    Highly efficient agentic 501B parameter open model

    #3 on Product Hunt

  • Umbra Mailnew

    Shows Gmail threads as chat conversations, a calmer interface over the same inbox.

    #4 on Product Hunt

  • Point and Yapnewcard

    Point at your screen and start yapping to your agent via MCP

    #5 on Product Hunt

  • Torinew

    Another AI support agent for customer success teams; the category is crowded and needs a sharper wedge.

    #6 on Product Hunt

  • Lazy AWSnew

    A keyboard-driven console for AWS resources, for people tired of the web console.

    #7 on Product Hunt

  • Wally by Willownew

    A writing agent that learns your style over time, betting on personal memory as the moat.

    #8 on Product Hunt

  • CoolDocknew

    A second Mac Dock with live widgets, a small utility in a well-worn category.

    #9 on Product Hunt

Show 41 more
  • Switch Labnew

    Lets keyboard buyers hear and type on switches virtually before buying, a niche but clear pain.

    #10 on Product Hunt

  • Quellonew

    One search box across your work tools, competing with enterprise search assistants.

    #11 on Product Hunt

  • Figma Specsnew

    Keeps design context and specs on the Figma canvas where agents and developers can read it.

    #12 on Product Hunt

  • AutoChart.eunew

    Turns songs into playable charts for rhythm games, a narrow hobby tool with a clear audience.

    #13 on Product Hunt

  • DoOnenew

    A focus timer tied to GitHub and GitLab tasks, connecting time tracking to real work items.

    #14 on Product Hunt

  • Starkomandnewcard

    Run your AI coding agents on an infinite canvas.

    #15 on Product Hunt

  • Framefieldsnew

    Code-first video creation tooling released as open source, aimed at programmatic and agent-made video.

    #16 on Product Hunt

  • Kai Resolve by Gleapnew

    A support assistant that investigates tickets before handing them to a human.

    #17 on Product Hunt

  • CrawlForge MCPnew

    Bundles 31 web tools into one MCP server so agents get scraping and search in one install.

    #18 on Product Hunt

  • Planet-Xnew

    A memorial globe for remembering loved ones, a consumer app outside the usual AI wave.

    #19 on Product Hunt

  • walkai-talkainew

    Talk to your coding agent over a walkie-talkie, an experiment in away-from-desk agent control.

    #20 on Product Hunt

  • Own 1new

    A personal AI that keeps your data under your control, pitching ownership as the feature.

    #21 on Product Hunt

  • buddylistnew

    Shows your coding agents as a 2005-style buddy list, a playful status view for parallel agents.

    #22 on Product Hunt

  • Triwardennew

    A native Mac client for Bitwarden and Vaultwarden, for users who dislike the web-based apps.

    #23 on Product Hunt

  • croftnew

    TypeScript data pipelines stored in a single DuckDB file, a lightweight alternative to heavy data stacks.

    #24 on Product Hunt

  • Beam.winnew

    Run agents on one Mac and use their windows from another, splitting compute from where you work.

    #25 on Product Hunt

  • readywhennew

    A personal assistant inside WhatsApp aimed at older users, meeting them in an app they already use.

    #26 on Product Hunt

  • Lancettanew

    Tells you when to push your coding agents harder and when to slow down, a usage-pacing tool.

    #27 on Product Hunt

  • XBOX Nano Controller

    A pocket-sized Xbox controller with full inputs, aimed at phone and handheld gaming.

    #28 on Product Hunt

  • PasteKit

    Formats JSON, YAML and SQL locally so pasted data never leaves your machine.

    #29 on Product Hunt

  • Gravity Notes

    One shared note list for you and your agents, treating notes as agent memory.

    #30 on Product Hunt

  • Inbox Zero

    An open-source email client positioned against Superhuman, with AI triage.

    #31 on Product Hunt

  • Tower 18 for Mac

    A major release of a long-running Mac Git client, now with better command-line support.

    #32 on Product Hunt

  • 3AM Super Intelligence

    Claims to investigate, prove and fix production incidents, a bold pitch in the AI on-call space.

    #33 on Product Hunt

  • Tomo for ChatGPT

    Turns ChatGPT answers into short Duolingo-style games, aiming at retention rather than answers.

    #34 on Product Hunt

  • Vampire Crawlers™

    A new game from the Vampire Survivors studio, ranking high on Product Hunt.

    #35 on Product Hunt

  • Latch

    An app blocker you unlock by walking to a physical key, adding friction on purpose.

    #36 on Product Hunt

  • Bee Frontier and Bee Eclipse

    Amazon's personal memory assistant comes to the wrist and phone, a big company bet on always-on recall.

    #37 on Product Hunt

  • shim-clicard

    See what your coding agent sends. Mask what you can.

    #38 on Product Hunt

  • Krea 3

    Krea's own image model, with 2K edits in under eight seconds as the headline claim.

    #39 on Product Hunt

  • Openbroca

    Lightweight dictation for Mac that adapts to your voice, one of several local speech tools today.

    #40 on Product Hunt

  • MeetPaw

    Starts meetings from the Mac menu bar, a small speed utility.

    #41 on Product Hunt

  • MyLens

    AI-driven one-to-one presentations at scale, pitched at sales and onboarding.

    #42 on Product Hunt

  • Popcraft

    Lets any agent generate video, images and audio through one tool.

    #43 on Product Hunt

  • SpeechFast

    Local Whisper transcription for Mac, another sign speech-to-text has moved on-device.

    #44 on Product Hunt

  • Spiralens

    A camera app that guides you to golden-ratio framing, a narrow but clear consumer idea.

    #45 on Product Hunt

  • Step 5 Preview

    A 600 billion parameter open-weight model trained for long-running agents, another entry in the open frontier race.

    #46 on Product Hunt

  • Salience

    Shows every request to your website and how it responds, request-level observability for small sites.

    #47 on Product Hunt

  • Rusteerio by Context.dev

    Keeps the Cheerio HTML parsing API but runs it on a faster Rust engine, an easy speedup for scrapers.

    #48 on Product Hunt

  • Craft Agents

    An agent app focused on user experience; the listing leans on superlatives without specifics.

    #49 on Product Hunt

  • LooGo

    Finds nearby public restrooms, a simple utility with daily demand.

    #50 on Product Hunt

Show HN

6

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

334 items from 77 sources, 8 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 TrendingMain12Fetched
Product HuntMain50Fetched
Show HNMain6Fetched
HN front pageMain30Fetched
YC LaunchesMain20Fetched
HF trending modelsMain20Fetched
HF trending spacesMain20Fetched
HF daily papersMain20Fetched
LobstersMain25Fetched
Dev.to topMain20Fetched
TechmemeMain15Fetched
AndroidX releasesAndroid · Official & releases14Fetched
Android Developers BlogAndroid · Official & releases4Fetched
Android Studio releasesAndroid · Official & releases2Fetched
Kotlin blogAndroid · Official & releases4Fetched
Kotlin releasesAndroid · Official & releases0Fetched
Compose Multiplatform releasesAndroid · Official & releases4Fetched
Android Developers YouTubeAndroid · Official & releases0Failed
Kotlin on YouTube (JetBrains)Android · Official & releases0Failed
Google Developers BlogAndroid · Official & releases1Fetched
AndroidX merged API changesAndroid · AOSP / AndroidX insider11Fetched
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 · Experts1Fetched
CommonsWare (Mark Murphy)Android · Experts2Fetched
TechYourChance (Vasiliy Zukanov)Android · Experts1Fetched
Kt. Academy (Marcin Moskała)Android · Experts2Fetched
Touchlab (Kotlin Multiplatform)Android · Experts2Fetched
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 releases0Fetched
Molecule releasesAndroid · Library releases0Fetched
Turbine releasesAndroid · Library releases0Fetched
Paparazzi releasesAndroid · Library releases0Fetched
SQLDelight releasesAndroid · Library releases1Fetched
OkHttp releasesAndroid · Library releases0Fetched
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 releases0Fetched
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 & ecosystem6Fetched
Android AuthorityAndroid · Leaks & ecosystem4Fetched
Android FaithfulAndroid · Leaks & ecosystem1Fetched
Android WeeklyAndroid · Community15Fetched
GitHub Trending · KotlinAndroid · Community12Fetched
Lobsters · androidAndroid · Community3Fetched
HN · AndroidAndroid · Community9Fetched
ProAndroidDevAndroid · Community1Fetched
Kotlin WeeklyAndroid · Community1Fetched
Philipp Lackner (YouTube)Android · Community0Failed
1 of 5

Gemini 4 Argon arrives for Android developers

Agents can now drive your local Android build and device tools through an official plugin, not only suggest code.

Google brought its newest model, Gemini 4 Argon, to Android development. You can use it in Android Studio Canary by signing into Antigravity, Google's agent app, or through the paid Gemini API. Google also shipped an Android CLI plugin that connects any terminal agent to your local Android tools. The Antigravity route needs a Google AI Ultra plan, and the post gives no benchmark numbers.

Android Developers Blog · OfficialRead the source

1 of 4

A local AI job search kit is GitHub's fastest riser

It shows that a well-packaged set of agent workflows can spread like a product without a server or a company behind it.

An open-source job search framework built on Claude Code gained almost four thousand stars in a day and now has about fifty eight thousand. It evaluates job postings, tailors your CV, writes cover letters and helps you prepare for interviews. Everything runs on your own machine, and the author encourages people to fork it and adapt it.

GitHub Trending · Fastest risingRead the source