Android Daily ·
Googlebook gets a sandboxed Linux terminal for agents and Android work
New here? How to read this issue
- 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.
- Stop at “You’re caught up”, or go deeper. Everything after it is the full list of what was collected today, folded away.
- 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.
Today’s Android picks
- #1Official
Googlebook gets a sandboxed Linux terminal for agents and developers
Why it matters Agents can get full admin rights on a Googlebook without putting the rest of the laptop at risk.
The Android Developers Blog announced a preview Linux terminal for Googlebook, Google's new laptop platform. Every command runs inside a Debian virtual machine, so AI tools can be given admin rights safely, and one click resets it while your files stay. Wireless ADB already works, and USB debugging of Android devices from inside the VM is planned.
- #2AOSP insider
Material 3 screen transitions with predictive back are coming to Compose
Why it matters Compose apps will soon get Material screen animations that also follow the back gesture, with no custom code.
Two new code changes in AndroidX, Google's open Jetpack libraries, add Material 3 shared axis and fade through transitions to Compose, Android's UI toolkit. Both support predictive back, where the previous screen peeks in as you swipe back. Teams that hand-built these animations can plan to switch once they ship in a release.
- #3Official
DataStore 1.3 reaches release candidate
Why it matters The stable release is close, so now is the time to try the new builder and built-in encryption.
DataStore, Jetpack's library for saving settings and small data, reached version 1.3.0 release candidate. The 1.3 line adds a new builder that replaces the old factory, an artifact for encrypted storage using Tink, support for web storage in Kotlin Multiplatform, and fixes for updates that were dropped during simultaneous reads and writes.
- #4Community
One tap on a web page could change call forwarding on some phones
Why it matters If your app can place calls, it must confirm before dialing carrier codes, because Android will not do it for you.
A security researcher showed that apps with the call phone permission can run carrier codes that change settings like call forwarding, with no confirmation. One popular dialer could be triggered from a web page with a single tap and is shipping a fix. Google closed the platform report as won't fix, saying it is up to apps.
- #5Official
JetBrains now recommends Kotlin Toolchain for new multiplatform apps
Why it matters New Kotlin Multiplatform projects may start without a hand-written Gradle setup.
JetBrains released Kotlin Toolchain version 0.13, a single entry point for building Kotlin apps for Android, iOS, desktop and servers using simple declarative configuration. With this release, JetBrains says it now recommends the toolchain for most new Kotlin Multiplatform projects, a shift away from starting every project with a custom Gradle build.
Also worth knowing
Releases, policy and leaks an Android engineer should know about this week.
- Adopt nowInsider
Runtime crashes hiding in your Gradle files - Nacho Carrión
Touchlab explains how transitive dependencies drift out of sync and cause runtime crashes, and how a Kotlin/Native change makes it worse.
Kotlin Multiplatform teams with many shared dependencies.
03 · Reference
Go deeper: every Android item, by lane
Official & releases
Google previewed a sandboxed Linux terminal for Googlebook, DataStore 1.3 reached release candidate, Tracing 2.0.4 is a safe stable update, and Compose 1.13 stays in beta.
AndroidX library releases
Stable · safe to ship · 5
Beta · API frozen, try it before it ships · 15
- Datastore 1.3.0-rc01
- Compose Animation 1.13.0-beta01
- Compose Foundation 1.13.0-beta01
- Compose Material 1.13.0-beta01
- Compose Material3 1.5.0-beta01
- Compose Runtime 1.13.0-beta01
- Compose Ui 1.13.0-beta01
- Core 1.1.0-rc01
- Graphics 1.1.0-beta01
- Heifwriter 1.2.0-rc01
- Palette 1.1.0-rc01
- Webkit 1.18.0-rc01
- Xr Arcore 1.0.0-rc02
- Xr Runtime 1.0.0-rc02
- Xr Scenecore 1.0.0-rc02
Alpha · experimental, APIs may change · 15
- Annotation Keep 1.0.0-alpha01
- Appfunctions 1.0.0-alpha13
- Benchmark 1.6.0-alpha01
- Car App 1.9.0-alpha03
- Metalava 1.0.0-alpha16
- Compose Material3 Adaptive 1.4.0-alpha03
- Compose Remote 1.0.0-alpha21
- Compose Remote Foundation 1.0.0-alpha04
- Core Uwb 1.1.0-alpha03
- Glance Wear 1.0.0-alpha20
- Navigation3 1.3.0-alpha02
- Test Backup 1.0.0-alpha02
- Wear Compose Remote 1.0.0-alpha13
- Webgpu 1.0.0-alpha06
- Xr Glimmer 1.0.0-alpha21
- Introducing the Googlebook Linux Terminal Preview: A VM-Backed Environment for AI and Developmentnewcard
Posted by Cyrus Stoller, Group Product Manager, Laptops & Tablets Today, getting the most out of AI workflows requires access to a terminal with a command line, even if someone doesn’t identify as a programmer or software developer. With Googlebook, we’re excited to help users…
Posted by Simona Milanovic, Developer Relations Engineer As Android developers, you have many choices when it comes to the agents, LLMs, tools, and command-line interfaces (CLI) you use for app development. Our goal is to help you build beautiful, high-quality Android apps, no…
Meta's Instagram Direct team on AI-native UI with Compose, cutting agent token cost per session by 33%; a real reference architecture.
Google Play's subscription changes for AI-era apps; read before pricing new features.
Newest Android Studio canary, for early adopters.
Previous Android Studio canary.
Android Studio Rabbit 1 is stable; time to update.
JetBrains' unified Kotlin Toolchain makes starting a KMP app simpler, with fewer Gradle decisions.
Show 6 more
KMP plugin drops Swift IDE features; iOS-side editing goes back to Xcode.
Planned extensions to Kotlin companion objects; language direction for multiplatform interop.
JetBrains' State of Kotlin 2026 report; handy adoption numbers.
Compose Multiplatform 1.13 alpha; web font and image preloading are no longer experimental.
Google's Developer Knowledge API and MCP server; official docs as machine-readable context for agents.
Google has officially released version 1.0 of the Agent Development Kit (ADK) for Kotlin, achieving full feature parity with the Python and Java ADK cores to enable idiomatic, multi-agent AI development. Built on Kotlin Multiplatform (KMP), the framework leverages Kotlin Symbol…
AOSP / AndroidX insider
Material 3 shared axis and fade through transitions with predictive back are landing in Compose, and AppFunctions annotations are still changing.
Remove identity hash from DynamicRange.toString()
CameraX DynamicRange now prints without an identity hash, a small cleanup for stable logs and tests.
[Tate] Add column metrics to VerticalTextLayout
Compose text gains column metrics for vertical Japanese and Chinese layout.
Add Material 3 shared axis transitions including predictive back
Material motion patterns are arriving in Compose Material 3 with predictive back built in, so hand-made transitions can go.
Remove isDescribedByKDoc default value from AppFunction annotations
AppFunctions, the API agents use to call into apps, drops a default, so descriptions must be set explicitly.
Add Material 3 fade through transition including predictive back
Material 3 fade through transition with predictive back is landing in Compose.
Remove ComposeUiFlags.isVelocityTrackerMinSampleSizeFixEnabled
A Compose fling velocity fix is now always on; scrolling may feel slightly different.
Add tip-of-tree dependency exemptions for wear.tiles and glance.wear
Wear Tiles and Glance Wear now build together at head, hinting at coordinated releases.
[owners] Remove sudheersai@google.com from appsearch/OWNERS
Routine AppSearch ownership change.
Show 1 more
[owners] Remove adorokhine@google.com from appsearch/OWNERS
Routine AppSearch ownership change.
Experts
Touchlab's post on runtime crashes from drifting Gradle dependencies is worth reading, and skydoves released Crayfish, an image cropper for Compose Multiplatform.
skydoves on what happens when an app opens a 108MP photo; practical OOM avoidance.
Crayfish, a new Compose Multiplatform image cropper from skydoves.
How Compose decides a class is unstable and what it costs; essential for performance work.
What the Compose compiler actually generates from your @Composable.
Balloon tooltips rewritten for Compose Multiplatform.
Haze 2.0 brings Glass, a new effects API, performance improvements, and experimental Android native backdrop support to Compose Multiplatform.
When Compose's new progressive blur replaces Haze and when it doesn't.
- Pixelating Composables in Compose Multiplatformin 2 sources
When it comes to displaying portfolio data inside of Sleevd, this is only available to users on the Ultimate plan. To hide this functionality from free users, a common approach here is to swap the value out for a placeholder or completely hide the component from view. While this…
Show 9 more
Joe Birch shows how to animate number changes in Compose, useful for finance and stats screens.
CommonsWare's weekly list of new Jetpack artifacts, handy for spotting new libraries early.
An earlier CommonsWare list of new Jetpack artifacts.
Reviews Shopify's claim that AI tips the balance from cross-platform back to native mobile code.
Kotlin's runCatching also catches cancellation; this post shows a safer version for coroutines.
A pattern for modelling loading, success and error states in Compose without boolean soup.
Touchlab's picks from next.app DevCon's 400 sessions.
Argues shared Kotlin code keeps fast, AI-assisted mobile teams from drifting between platforms.
Explains runtime crashes caused by mismatched transitive dependencies in multiplatform builds.
Library releases
SQLDelight 2.4.1, Roborazzi 1.76.0 and Haze 2.0.1 are small, safe patch updates with no breaking changes noted.
- Haze 2.0.1new
Patch release for Haze 2.0, the Compose blur library.
SQLDelight patch that speeds up code generation and documents Android transactions.
Roborazzi fixes duplicated options when rebuilding Compose screenshot test settings.
Leaks & ecosystem
Android 17 is reaching Nothing Phone 3 and older Galaxy phones, and leaks point to Gemini shopping overlays and Photo Picker editing.
Gemini may let users publish their creations as reusable styles, a step toward a creator marketplace.
Gemini's overlay may add a Shop this action, pushing shopping into the system layer above apps.
Gmail's Help Me Write may get a faster entry point.
- Google Translate could soon auto-save transcripts of your Live conversations (Updated: Rolling out)new
Live Translate transcripts can now be saved on mobile.
Gemini Live tests a new waveform and a voice output control.
New Gemini voices coming to Gmail Live; consumer feature, low developer impact.
Android may let a gamepad act as a mouse; an accessibility input path worth testing against.
Photo Picker may gain flip, rotate and crop; could remove your own crop step.
Show 12 more
Monthly Play services and Play Store changes, worth scanning for API and policy updates.
Daily list of discounted Play Store apps and games.
One UI 9 with Android 17 reaches Galaxy S24 and Fold and Flip 6 in the US; test your app on it.
9to5Google's review weighs whether Googlebook can become a real PC platform.
Nothing Phone 3 gets stable Android 17.
LineageOS 24 brings Android 17 to custom ROM users with a simpler flashing tool.
Pixel 11's Rambler voice typing has been ported to other phones.
Googlebook's Android 'Better Together' sync expands, including to Samsung phones.
Meta's Wonder models are its bid to make the Muse assistant competitive.
Nothing OS 5.0 on Android 17 is rolling out to Phone 3.
Samsung's Android 17 rollout widens to more US Galaxy phones.
Podcast hands-on with the HP Googlebook 14.
Community
Dialer app makers should read the one-tap MMI code issue, and LineageOS 24 brings Android 17 to custom ROM users.
Hey HN, I’m Vishnu from Agent37 ( https://agent37.com ). We build sandboxes for running AI agents at under 10% of the cost of most sandbox providers. Our customers spend an average of $0.37/month per sandbox. You can view a video walkthrough of the product here…
Show 38 more
Open-source download manager in Kotlin for desktop and Android.
Hook-based location-spoofing app trending; policy-sensitive territory.
Kernel-based root solution, trending alongside the new CVE.
A local and YouTube Music player for Android.
Android GUI for the sing-box proxy.
Weather app aggregating 50+ sources with strong visualisations.
A YouTube Music client styled after Apple Music.
Minecraft Java Edition launcher for Android.
V2Ray client for Android.
Minimal launcher aimed at cutting screen time.
One-tap exploit app for CVE-2026-43499; check your devices' security patch level.
KernelSU fork with kernel patch modules.
- ahXN00/OwnTVnew
Native IPTV player for Android TV built with Compose for TV, ExoPlayer and mpv.
Runs PC games from Steam, Epic and GOG on Android.
Media streaming app for Android.
Cross-platform YouTube Music client.
Open-source manga reader for Android.
LineageOS 24.0 is out; the custom-ROM community tracks Android 17 quickly.
One-click MMI code execution on Android; a security issue worth reading.
Large debate on whether Linux phones can break the Android and iOS duopoly.
Policy discussion on affordability, not Android-specific.
Map of Google's data centers worldwide.
An AI agent on a 48MB-RAM Nokia feature phone; agent UX on minimal hardware.
Discussion of system-level ad blocking on Android.
Walks through Flow collection in Compose up to lifecycle-aware collection.
Explains Kotlin's planned alternative to companion objects for static members.
Case study on fixing dependency conflicts when moving an Android app to Bazel.
Explains SHA-1 and SHA-256 where Android devs meet them: signing, APIs and Firebase.
Benchmarks Room full-text search against SQL LIKE for in-app search.
Kotlin Weekly issue 532, led by the end of Swift editing in the KMP plugin.
Today’s Product & AI picks
- #1In 2 sources
Agent37 rents agent sandboxes for about 37 cents a month
Why it matters If each user gets their own agent computer, cheap isolation decides whether the product makes money.
Agent37, a new Y Combinator company, launched on Hacker News with sandboxes for running AI agents at under a tenth of the cost of most providers. The founders say customers spend about 37 cents a month per sandbox on average. It joins Maritime, another launch this week that gives every agent its own persistent virtual machine.
- #2Worth a debate
Codex CLI quietly runs a background daemon, and developers push back
Why it matters Agent tools that keep running after you close them will lose trust unless that behaviour is opt-in.
A bug report on OpenAI's Codex command line tool reached the top of Hacker News with over 500 points. The reporter found it starts a persistent background service by default, copies a package of nearly 400 megabytes into the home folder, and updates itself, which can interrupt running work. Commenters asked for it to be opt-in.
- #3New idea
Janas runs an 80 billion parameter model on a 32 gigabyte laptop
Why it matters Large open models may run on ordinary laptops by keeping most of the model on disk.
Janas is a new open-source engine written in plain C. It runs mixture of experts models, which use only a few parts per word, by streaming those parts from a fast SSD as needed. The author reports an 80 billion parameter model running on a 32 gigabyte laptop, many times faster than llama.cpp, on their own test machine.
- #4In 3 sources
A music workstation lets AI agents drive it through MCP
Why it matters Adding an MCP bridge lets agents work inside an existing creative app instead of replacing it.
Blokkily, shown on Hacker News, is an experimental music workstation built around patterns, with a tracker, piano roll and step grid. It runs on Linux, Windows and Apple Silicon Macs. An MCP bridge, the common plug for AI tools, lets an agent control the running app, and a demo shows one composing a chiptune.
02 · Reference
Go deeper: themes, product notes, every source
Themes across sources
Agent sandboxes · MCP-native apps · On-device AI · Coding-agent control surfaces · Agent-ready Android
Agent sandboxes
Several launches sell the same thing: an isolated computer for each AI agent. Agent37 competes on price, Maritime on scale, and Google gives Googlebook a VM terminal that agents can control. On the other side, Codex CLI's always-on background daemon drew a large debate about what agent tools may run on your machine.
Why now. Agents now run for hours with shell access, so isolation and cost per agent have become product decisions, not afterthoughts.
Launch HN: Agent37 (YC F26) – Low-cost agent sandboxes for running agents
MCP-native apps
Show HN and Product Hunt carry apps that expose their own controls to agents through MCP: a music workstation, a CAD tool, a driving journal, a diagram tool and marketing data. Others rebuild web basics, forms and page fetching, for agents as users.
Why now. MCP has become the standard plug, so adding agent control to an app is now a small feature, not a platform bet.
Show HN: An experimental pattern-first DAW that supports MCP
On-device AI
Big models keep moving onto small hardware. Janas streams model experts from an SSD to run an 80 billion parameter model on a laptop, AirLLM and BitNet trend on GitHub, and EmbeddingGemma 2 lands on Product Hunt with browser and GGUF builds.
Why now. Sparse expert models, 1-bit weights and fast SSDs together cut the memory a useful model needs.
Janas: Run LLMs larger than RAM by streaming experts from SSD
More
Coding-agent control surfaces
Tools for watching and steering many coding agents keep shipping: terminals that flag the waiting agent, notch panels, an infinite board for agents, and a voice assistant built on Claude in print mode.
Why now. Developers now run several agents at once, so attention, not typing, is the bottleneck.
Agent-ready Android
Google keeps feeding agents official Android knowledge and hooks: the Developer Knowledge API and MCP server, the Android CLI's skills, ADK for Kotlin, and AppFunctions annotations still changing in Gerrit.
Why now. Coding agents write much of today's Android code, so Google wants them grounded in current guidance.
Supercharge your development with the Google Developer Knowledge API ecosystem
More
- Device Streaming and Android skills - available in Android CLI
- Announcing ADK for Kotlin 1.0: Building Production-Ready AI Agents in Kotlin, Android, and Beyond
- How Instagram Direct engineers built AI-native UI architecture with Jetpack Compose and reduced token cost per agent session by 33%
- Appfunctions 1.0.0-alpha13
- Remove isDescribedByKDoc default value from AppFunction annotations
- Harness Engineering: What I Learned From 24,000 Prompts Building Production Apps
Product notes
Who each product is for, the problem, why now, its wedge and how it makes money. Today’s picks first.
Launch HN: Agent37 (YC F26) – Low-cost agent sandboxes for running agents
- What
- Agent37 rents sandboxes for running AI agents at under 10 percent of the cost of most sandbox providers, about 37 cents a month per sandbox on average.
- For whom
- Teams running many always-on or per-user agents where sandbox cost dominates.
- Problem
- Per-agent isolated environments are priced for short jobs, so running one per user gets expensive fast.
- Why now
- Agent products now give each user a persistent computer, which turns sandbox cost into a unit-economics problem.
- Wedge
- Price: a tenth of incumbents for the same basic isolation.
- Money
- Usage-based sandbox hosting.
- Like
- E2B · Modal · Daytona · Maritime
LessonWhen a new workload makes an old cost line huge, a 10x cheaper version of the same primitive is a full wedge.
Janas: Run LLMs larger than RAM by streaming experts from SSD
- What
- Janas, a dependency-free C engine that runs mixture-of-experts language models larger than RAM by streaming expert weights from an SSD.
- For whom
- Developers and hobbyists who want large open models on an ordinary laptop, with or without a GPU.
- Problem
- Large models need more memory than laptops have, so people fall back to small models or the cloud.
- Why now
- Mixture-of-experts models only use a few experts per token, and NVMe SSDs are fast enough to fetch them on demand.
- Wedge
- Runs an 80 billion parameter model on a 32 gigabyte laptop, per the author's own benchmark.
- Money
- Open source under GPL, no model yet.
- Like
- llama.cpp · AirLLM · Apple LLM in a flash
Speed figures come from the author's single-machine benchmark, and the project calls itself experimental.
LessonLook at which resource a design wastes; here most model weights sit idle each token, so they can live on disk.
Show HN: An experimental pattern-first DAW that supports MCP
- What
- Blokkily, an experimental pattern-first music workstation for Linux, Windows and Apple Silicon Macs, with an MCP bridge that lets agents control the running app.
- For whom
- Electronic musicians and hobbyists who work in patterns, and people curious about agent-assisted composition.
- Problem
- Desktop creative apps are closed to agents, so AI music help lives in separate generators.
- Why now
- MCP gives any desktop app a cheap, standard way to expose its controls to agents.
- Wedge
- Agent control of a real editing tool, shown with a chiptune demo, instead of one-shot song generation.
- Money
- Open source hobby project, no model yet.
- Like
- Renoise · LMMS · Suno
LessonExposing your app's own controls to agents keeps the human in the editor rather than replacing the editor.
Introducing the Googlebook Linux Terminal Preview: A VM-Backed Environment for AI and Development
- What
- A preview Debian Linux terminal on Googlebook. Every command runs inside an isolated virtual machine, and AI tools can be given admin rights inside it.
- For whom
- Googlebook owners who use AI agents or command-line tools, from non-programmers to web and Android developers.
- Problem
- Running an agent with full shell access on your main system is risky, and Googlebook had no native command line for these workflows.
- Why now
- Agents now do real work in a terminal, so a laptop platform without a safe shell loses developers.
- Wedge
- A sandboxed VM that agents can break and reset in one click while host files stay safe.
- Money
- Platform feature that sells Googlebook hardware; no direct charge.
- Like
- ChromeOS Linux (Crostini) · WSL · Docker Desktop
LessonWhen users will hand power to an agent, ship the sandbox and the reset button with the feature, not after the first incident.
mksglu/context-mode
- What
- An open-source layer for coding agents that sandboxes tool output, keeps session memory, and routes tools across 17 agent platforms through MCP and hooks.
- For whom
- Developers whose coding agents burn their context window on long logs and tool dumps.
- Problem
- Tool output fills the context window, so agents forget earlier work and cost more per session.
- Why now
- MCP and hook systems in coding agents now give a standard place to intercept tool output.
- Wedge
- A claimed 98 percent cut in tool output sent to the model.
- Money
- Open source, no model yet.
- Like
- Repomix · Claude Code hooks · Serena
The 98 percent reduction is the project's own figure and depends heavily on the workload.
LessonSit between the agent and its tools: the plumbing layer can win across every agent at once.
Show HN: SolidCraft – Open-Source Parametric CAD in Rust (With MCP Support)
- What
- SolidCraft, an open-source parametric CAD tool written in Rust with MCP support.
- For whom
- Makers and engineers who model parts in code and want agents to edit models.
- Problem
- Parametric CAD is mostly closed, and agents cannot drive it.
- Why now
- Agents are good at editing code-like models, and MCP makes the hook-up standard.
- Wedge
- Open source plus agent access, a combination big CAD vendors do not offer.
- Money
- Open source, no model yet.
- Like
- OpenSCAD · CadQuery · FreeCAD · Zoo
LessonIn a market owned by heavy incumbents, an open, agent-drivable core is a credible side door.
Cube Forms
- What
- Cube Forms, a form builder whose forms can be filled in by AI agents as well as people.
- For whom
- Teams collecting sign-ups, orders or applications that expect agents to act for users.
- Problem
- Web forms are built for human clicks, so agents fill them unreliably through the browser.
- Why now
- Agents now act on users' behalf, and form owners want those submissions to arrive clean.
- Wedge
- Agent-readable forms, a single feature incumbents lack today.
- Money
- Likely subscription SaaS like other form builders.
- Like
- Typeform · Tally · Google Forms
The listing gives no detail on how agents fill the forms or which agents are supported.
LessonLook for everyday web objects that assume a human is clicking; each one is a candidate for an agent-native rebuild.
Expand Fetch: Reading glasses for your AI agent
- What
- Expand Fetch takes a URL and returns accurate Markdown, the page's hidden JSON, and a snapshot an agent can cite.
- For whom
- Developers building agents and retrieval systems that read live web pages.
- Problem
- Scraped pages come back noisy or incomplete, and agents cannot prove where a fact came from.
- Why now
- Agents now browse at scale, and citations have become a product requirement.
- Wedge
- Citable snapshots, not just cleaner text.
- Money
- Usage-based API, YC-backed.
- Like
- Firecrawl · Jina Reader · Context.dev
LessonIn a crowded API category, pick the one output buyers must defend later, here citations, and make it the headline.
Introducing Crayfish: An Image Cropper for Compose Multiplatform
- What
- Crayfish, an image cropper library for Compose Multiplatform with a one-line crop dialog, a Cropper composable, aspect ratios, gestures, shapes and Coil support.
- For whom
- Compose Multiplatform developers who need profile-photo or document cropping on Android and iOS.
- Problem
- Image cropping on Android leans on older View-based libraries that do not work on other platforms.
- Why now
- Compose Multiplatform is stable on iOS, so teams want shared UI pieces that used to be platform-specific.
- Wedge
- A one-line dialog for the common case, plus a composable for custom layouts.
- Money
- Open source, part of the author's library portfolio and paid newsletter.
- Like
- uCrop · Android-Image-Cropper · Compose Cropper
LessonShip the one-line default first and the flexible building block second; most users only ever need the first.
morluto/rea
- What
- An open-source tool that uses AI agents to reverse-engineer software, from observed app behaviour down to native binaries.
- For whom
- Security researchers, modders and engineers who need to understand closed software without source.
- Problem
- Reverse engineering is slow, expert-only work in tools like Ghidra or IDA; most developers cannot do it at all.
- Why now
- Coding models can now read disassembly and iterate; the same wave is decompiling old games at scale (see Techmeme today).
- Wedge
- One open repo with a sweeping promise ('reverse engineer anything'), riding agent interest to viral GitHub growth.
- Money
- None stated; open source. A hosted or enterprise security product is the obvious later path.
- Like
- Ghidra · IDA Pro · Binary Ninja
'Anything' is a very broad claim, and ~25k stars in a day reflects virality more than proven results on real binaries.
LessonAn audacious promise on a hot capability buys distribution; retention depends on whether it survives contact with real targets.
Maritime: Fully managed cloud for running AI agents at scale
- What
- A managed cloud where every AI agent runs on its own isolated, persistent VM.
- For whom
- Companies running many long-lived agents that need state, tools and isolation.
- Problem
- Serverless functions and shared containers fit badly with agents that run for hours, keep state and execute untrusted code.
- Why now
- Agents moved from chat turns to long-running jobs with real tool access, which needs persistent and safe compute.
- Wedge
- One infrastructure primitive, a VM per agent, promised at 'millions of agents' scale.
- Money
- Usage-based compute pricing, like other cloud platforms.
- Like
- E2B · Modal · Fly Machines
LessonA new workload shape (long-lived, stateful, untrusted) can open space for new infrastructure even in a mature cloud market.
Farol
- What
- A macOS terminal that shows which of your coding agents needs your attention.
- For whom
- Developers running several coding agents in parallel.
- Problem
- With multiple agents running, it is easy to lose track of which one is blocked waiting for input.
- Why now
- Running several agents at once is becoming normal; Littleguys, Halo and Agibot launched around the same idea today.
- Wedge
- Owning the terminal, where coding agents already run, instead of adding a separate dashboard.
- Money
- Not stated; likely a paid Mac app or a pro tier.
- Like
- Littleguys · Warp · iTerm2
LessonWhen users shift from doing to supervising, the valuable UI is attention routing, not more output.
google/embeddinggemma-2
- What
- Google's EmbeddingGemma 2: an Apache-2.0 multimodal embedding model covering text, images, audio and video.
- For whom
- Developers building search, RAG and recommendations, including on-device and in-browser.
- Problem
- Strong multimodal embeddings have mostly been API-only or too large to run locally.
- Why now
- Within days it is quantised (Unsloth GGUF) and running in the browser over WebGPU, showing small open embeddings can be deployed anywhere.
- Wedge
- Open licence plus Google's name, small enough for devices.
- Money
- Free; it strengthens the Gemma ecosystem and Google Cloud.
- Like
- OpenAI text-embedding-3 · Jina Embeddings · Nomic Embed
LessonOpen, small and multimodal is a distribution strategy: the community ships the ports for you.
Device Streaming and Android skills - available in Android CLI
- What
- Android CLI adds Device Streaming (drive remote physical devices) and 'Android skills' that ground coding agents in official guidance.
- For whom
- Android developers working with AI agents and command-line workflows.
- Problem
- Agents generate outdated Android code and cannot test on real hardware.
- Why now
- Coding agents are mainstream in Android teams; Google is making its docs and device fleet agent-accessible.
- Wedge
- Official, CLI-first tools any agent can call, not tied to one IDE.
- Money
- Free; it keeps developers on Google tooling and its device-testing infrastructure.
- Like
- Firebase Test Lab · Gemini in Android Studio · Developer Knowledge MCP
LessonPlatform owners win the agent era by making their ground truth (docs, devices) machine-readable.
Haze 2.0
- What
- Haze 2.0 by Chris Banes: a Glass effect, a new effects API, performance work and experimental native Android backdrop support for Compose Multiplatform.
- For whom
- Compose and Compose Multiplatform developers building blurred or glass-style UI.
- Problem
- Background blur is hard to do well and fast across platforms in Compose.
- Why now
- Glass-style UI is in fashion and Android now offers a native backdrop API Haze can sit on.
- Wedge
- One API across Android, iOS and desktop.
- Money
- Open source; reputation and community.
- Like
- Compose progressive blur · Cloudy · SwiftUI materials
LessonLibrary authors stay relevant by adopting new platform APIs underneath while keeping their own API stable; his companion post on progressive blur shows how to explain that.
By source
Everything 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
Reverse engineer anything with agents, from app behavior down to native binaries.
A spy satellite simulator in your browser, except the data is real. Live open source spatial intelligence on a photorealistic 3D globe.
Cuts tool output sent to coding agents and keeps session memory across 17 platforms.
An open 'crafting engine' for artists who want control over AI image and video, not one-shot prompts.
The popular self-hosted chat interface for local and API models keeps trending.
Editorial diagram design for Claude Code, Codex, GitHub Copilot, Cursor, Factory Droid, and Pi. 44 diagram types. Self-contained HTML + SVG. No shadows. No Mermaid slop.
Runs 70B models on a 4GB GPU by loading layers one at a time: speed traded for access on cheap hardware.
Microsoft's inference framework for 1-bit models, which run on ordinary CPUs.
Show 1 more
Routes many coding agent tools through free model quotas; check terms before relying on it.
Product Hunt
Notes that tell the tale
- Hexis Cloudnew
Shared memory for all of a company's AI agents.
AI-Powered LinkedIn Platform for Content and Outreach
A terminal coding agent that claims to learn from your sessions.
- Codemindnew
A coding agent you delegate whole tasks to.
Turns any website into clean data through one API call.
AI reminder scanner for bills, invites & documents
- Ancher 2.0new
Turn everything you read and save into work you can ship
Show 42 more
- Kudos 3.0new
The AI that finds money you're missing and gets it back
Forms built so AI agents can fill them in reliably.
- SalesRep.comnew
Your AI sales rep: live prices, quotes & follow-ups 24/7
- ngram 2.0new
Create explainers, screencasts, and AI avatar videos
- ExcaliClawnew
Hand-drawn diagrams an AI agent can draw with you.
- Milepostnew
An iOS driving journal that exposes trips to agents through MCP.
- SquadRoomnew
AI employees that work 24/7
- Manifold MCPnew
Marketing and SEO data packaged as an MCP server for agents.
- guilds.runnew
Builds multi-agent workflows with checks for trust.
EmbeddingGemma 2 lands on Product Hunt after topping Hugging Face.
- Base Codenew
Lets non-engineers build on a product's codebase with agents.
Volume control from the trackpad edge; tiny utilities still launch well on Product Hunt.
Free floor plans from a sketch or a prompt; generative layout for non-designers.
One of several notch utilities today; the MacBook notch has become a micro-platform.
Google Japan's annual joke keyboard; brand marketing more than product.
Japanese listening practice generated from your own sentences; personal-content language learning.
The macOS terminal that shows which coding agent needs you
AI swing coach for pickleball; vertical coaching riding a fast-growing sport.
Native, open-source Mac spreadsheet; a bet that people want small native apps over web suites.
Minecraft server directory with spoof detection; trust signals differentiate a listings market.
Cocktail finder from what's in your bar; the classic 'make it from what I have' utility.
Voice to-do list that asks before acting; 'ask first' as a trust feature for agents.
Always-on Mac agent, free forever; price is the wedge and monetisation is unclear.
Software room correction for Mac audio; DSP replacing acoustic panels.
Use any model inside Claude Code, keep your Claude plan
Split-flap message board for any TV; nostalgic design is the whole product.
Local-first writing app; offline and privacy as positioning in a saturated category.
Notch app bundling notifications, dictation and agent status; the notch as an agent dashboard.
Launch platform selling SEO plus 'AEO' (optimising for AI answer engines), now a product category.
A browser extension that turns browsing into a Pokédex hunt; engagement through collection, but IP-fragile.
Sales workspace pitching one shared context across tools: the 'context layer' idea applied to sales.
Replies to fans in a creator's voice; scales engagement but risks an authenticity backlash.
Perplexity adds recurring background automations; assistants are becoming schedulers.
A personal AI computer that runs entirely on-device
An agent that owns a team's meeting agenda; a painful but crowded wedge.
Job-search copilot whose angle is recruiter input rather than generic résumé rewriting.
Screen recorder where an agent does the edits; editing becomes an instruction, not a timeline.
Terminal with a project-aware Run button; IDE conveniences moving into the terminal where agents live.
- Hypervibenew
Run and watch many coding agents on one canvas.
Windows Search now takes actions, not just finds files.
Local AI chat, transcription and video tools in one Mac app
Record, edit and share video entirely in the browser; Loom's space with a no-install angle.
Show HN
Static recompilation runs PS2 games in the browser, the top Show HN today.
A 9th grader's hold-to-talk Mac assistant that hands tasks to Claude.
A multiplayer brick-building world that runs in the browser.
Claims a near-complete Figma clone in 8 prompts; a single anecdote, not a benchmark.
Open-source parametric CAD in Rust that agents can drive through MCP.
A personal site rewritten in server-side Swift.
HN front page
- Muse Wondernew
Meta's new Muse Wonder models top Hacker News.
Attackers switched supermarket fridges to defrost, a stark lesson in connected device security.
Codex CLI starts a self-updating background service by default, raising trust questions for agent tools.
- GLM-Legalnew
A legal-domain model from the GLM family draws a big discussion.
Builds a modern web app in plain C, arguing for fewer layers.
Argues that getting value from AI tools is a skill that must be practised.
Turns a WiFi 7 router into a software defined radio receiver.
Citroen brings back the 2CV as an electric car.
Show 22 more
Science story on the genetics of orange cats.
Archive interview on how Super Mario 64 was made.
Essay on how much single cells can compute.
Revisits Kevin Kelly's 2016 AI predictions a decade later.
Tracks down who named an obscure x86 instruction.
Plug-in balcony solar panels are spreading.
Intel's AI accelerator with up to 480 GB of memory targets local inference.
How Firefox made large downloads resumable.
Study rejuvenates blood cells into young stem cells.
Hey HN, I’m Vishnu from Agent37 ( https://agent37.com ). We build sandboxes for running AI agents at under 10% of the cost of most sandbox providers. Our customers spend an average of $0.37/month per sandbox. You can view a video walkthrough of the product here…
China passed the US in R and D spending in 2024.
A hobby internet phone switch.
Research argues models can provably hide capabilities, a challenge for safety testing.
How a tiny demoscene intro was written for the ZX Spectrum.
Examines OpenAI's claim of faster Fourier transform math.
A multiplayer brick-building world that runs in the browser.
History of the Mac Dock.
An automated research claim of integer multiplication below n log n.
Open-source parametric CAD in Rust that agents can drive through MCP.
Vi Hart's long explainer on the Navier-Stokes problem.
YC Launches
Agents for go-to-market work, pitched as Cursor for sales and marketing.
Returns clean Markdown plus a citable snapshot of any page for agents.
Finds repetitive work in a company and suggests what to automate.
Run millions of agents, each on its own isolated, persistent VM.
Voice agents interview employees to capture tacit process knowledge: a context layer for automation.
Agents handling onboarding through renewal; post-sale customer success as a vertical agent market.
Desktop harness that steers 100+ image and video models; orchestration over generation.
Physically synthesises AI-designed molecules; the bottleneck moves from design to the lab.
Show 12 more
Fraud and risk decision model claiming 1/100th the cost of a frontier model; decision models in fintech.
AI-native workers' comp administration; agents for claims-heavy insurance work.
'Codex for electronics': AI-native PCB design.
An AI chip pitched for recursive self-improvement; a very large claim at a very early stage.
Helps companies build their own in-house model lab; a bet that firms want their own models.
Parallel dialling inside Attio and GoHighLevel; sales tools living inside the CRM.
AI process consultant that maps workflows before automating; discovery before deployment.
AI for investor relations; a narrow, high-value, regulated workflow.
Captures unwritten business rules to guide AI rollout; the same context-layer bet as Sona8.
Health-assistant robots for ageing parents; eldercare as a robotics wedge.
A to-do list that does the tasks itself; the agentic to-do pattern, like Foley.
AI experiment design for biotech teams.
HF trending models
A model that makes AI text read as human-written; detection-evasion demand is real and contentious.
Faster Qwen-Image 2.1 variant; Qwen dominates today's HF columns.
Cloudflare publishing its own models; edge providers moving up the stack.
Lightricks updates its open video model; open video generation keeps pace.
Community fine-tune on a 27B base.
Mid-size Qwen release; 27B is the sweet spot for local use and fine-tuning.
A 3B model from Liquid AI; on-device class.
Show 10 more
Community model trending; details are thin.
Kitchen-sink merge with every frontier model in its name; treat such claims cautiously.
EmbeddingGemma 2 quantised within days; fast ecosystem uptake.
29B mixture-of-experts with 4B active, packaged for local use.
Community model trending; details are thin.
From Cactus, an on-device inference company; small models for phones.
Aleph Alpha's new model; the European sovereign-AI angle.
A small 3 billion parameter model trending on Hugging Face.
Faster tier of Cloudflare's new model.
Baidu's OCR model aimed at long documents.
HF trending spaces
A decision-model leaderboard; evaluation follows the decision-model wave.
Video generation demo built on Wan.
Tracks which models are gaining popularity; meta-tools for following the model firehose.
RL training across multiple agent harnesses; teaching agents to work in many tool setups.
Viggle's turbo Qwen-Image space; a company using HF as a distribution channel.
Experimental fast image-edit pipeline.
EmbeddingGemma 2 running in the browser via WebGPU; embeddings without a server.
Character-swap LoRA demo.
Show 6 more
Turbo variant demo of Krea 2.
Hybrid Qwen and Flux image pipeline experiment.
Video generation with automatic prompting.
An open recreation of a decision index demo.
Image-to-video demo.
HF daily papers
Probes how multimodal models represent space internally.
Humanoid robot control that reacts to what it sees.
Synthesizes software environments at scale to train computer-use agents.
Tests whether agents can learn strategies over a long game competition.
Keeps small models learning new skills without forgetting old ones.
Turns recordings of people using software into reusable agent skills.
A fast visual grounding model for robots.
Training data for long terminal tasks by agents.
Show 12 more
Attention Residuals (AttnRes) enables content-dependent retrieval over network depth, which can be viewed as single-stream softmax attention along the depth axis. Inspired by the success of multi-head attention along the sequence axis, we introduce Multi-Stream Attention…
Speculative decoding accelerates large language model inference by using a lightweight drafter to propose multiple tokens that are subsequently verified in parallel by the target model. Recent diffusion-based drafters make this paradigm particularly attractive: deep…
Benchmarks whether coding agents can build the agent a user actually wants.
Humans perceive far more in a scene than what is explicitly depicted: a single glance captures past causes and future trajectories; a quick peek determines if a vehicle can fit between two parked cars; a few seconds of video reveals who holds authority in a room; and a fleeting…
Finds offline distillation often beats on-policy distillation.
Recent GPU kernel optimization agents retain what they learn in knowledge bases or as distilled skills. Kernel benchmarks score each attempt's implementation for correctness and speed but leave the reuse value of retained experience unmeasured. We introduce MetaKernelBench…
Autoregressive (AR) grounding models serialize spatial predictions, introducing sequential latency and imposing a causal order on output tokens. We view grounding as visual evidence extraction: objects, locations, and spatial relations are jointly constrained by the image and…
When solving complex spatial reasoning tasks, ordinary humans naturally revisit the scenes from task-relevant viewpoints and mentally reconstruct the spatial layouts from those viewpoints. But can Large Vision-Language Models (LVLMs) proactively think in the same way? We term…
Data video is a media form that integrates data visualization with video narrative, widely adopted in news reporting and business analysis. Compared with general video understanding, data video understanding places greater emphasis on accurately reading data from animated…
Decoder hidden states from different layers of large language models (LLMs) often encode complementary information. However, existing speculative decoding methods typically rely on a small set of target-model layers chosen using fixed, position-based heuristics. Selecting an…
Uses 4,782 real JetBrains sessions to argue agent benchmarks should match real workflows.
Large-scale egocentric human interaction data is becoming an important source of physical supervision for embodied learning, yet video alone leaves the contact and pressure that characterize physical interaction unrecorded. Recent visual-tactile datasets provide this missing…
Lobsters
Essay on turmoil in the Ruby community.
Linus Torvalds used Claude Code to explore a kernel performance change.
Updated crypto recommendations for the post-quantum era.
Vi Hart's long explainer on the Navier-Stokes problem.
Essay arguing AI should not be treated as a neutral tool.
Write-up of a 90 thousand dollar remote code execution bug in Google.
What are you doing this week? Feel free to share! Keep in mind it’s OK to do nothing at all, too.
Browser fingerprinting is replacing cookies for tracking.
Show 17 more
French supermarkets confirm a cyberattack shut down fridges.
Unreal editor got faster by dropping overly secure hashing.
How a tiny demoscene intro was written for the ZX Spectrum.
Parsing JSON at tens of gigabytes per second from Python.
Tracks down who named an obscure x86 instruction.
Reminder that AI coding still requires your own thinking.
Builds a modern web app in plain C, arguing for fewer layers.
- D-Bus Hellnew
A popular rant on the complexity of D-Bus on Linux.
Primer on arena allocators in C.
- DataCube: A Relational Aggregation Operator Generalizing Group-By, Cross-Tab, and Sub-Totals (1997)new
Abstract: Data analysis applications typically aggregate data across many dimensions looking for anomalies or unusual patterns. The SQL aggregate functions and the GROUP BY operator produce zero-dimensional or one-dimensional aggregates. Applications need the N-dimensional…
A well-known blogger explains why they stopped posting.
Dev.to top
A 15-second screenless cognitive reset dispatcher combining deterministic NOAA solar ephemeris, 432Hz Web Audio synthesis, 1080p canvas field cards, and Entire AI provenance.
A few months ago, I published You're a Real JavaScript Developer Only If... It was supposed to be a...
- Solodesk-Ainew
What I Built SoloDesk AI is a privacy-first, zero-marginal-cost financial copilot and invoice...
Let me ask you something. What are you supposed to learn when the CEOs of all AI companies keep...
Hello, I'm Maneshwar, and I'm building LiveReview — a blast-radius aware AI code review built for...
Compares small open models against Gemma 4 on one TPU.
A single log line cost 40 thousand dollars a year in latency.
I tested Traefik 3.7's new service-level middlewares on a 90/10 weighted canary service: 20 of 200 requests landed on the canary branch and all 20 carried the canary-only header, with zero leaks across 500 concurrent requests. Traefik's own announcement blog shows YAML that…
Show 12 more
Local embeddings in Dart with llamadart, relevant for Flutter apps.
A step by step guide to Debian sid with kernel 7.2 and GNOME on the Snapdragon 7c Gen 2 Galaxy Book Go, a laptop with no upstream device tree. Firmware, interrupts, audio calibration and the charger protocol taken from Samsung's Windows install, 26 boots over two days, and an…
Lessons on AI slop from 9,050 generated scripts.
🎃 The October We Almost Burned Out Last Hacktoberfest, our repo (a mid-sized CLI tool with...
This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass CHECK OUR...
What it is, what it isn't, and how to do it
Weighs the tradeoffs in Torvalds' Claude-assisted Linux patches.
Argues agents cannot judge what is acceptable to break, so humans must.
Argues the usual benchmark for local Claude Code replacements is the wrong one.
Summarises Anthropic's talk on coordinating agents with Unix ideas.
Hello, I'm Rijul, and I'm building LiveReview — a blast-radius aware AI code review built for your...
One MCP server driving six AWS inference backends.
Techmeme
Chip stocks fall after Broadcom's CEO compares AI investing to a casino.
Character.AI files for bankruptcy and sells some assets to Disney.
Nvidia's AI chip shipments in China fell sharply under export curbs.
Transsion, Africa's leading phone maker, raised 390 million dollars in Hong Kong.
Anthropic says new safeguards cut distillation data harvesting by 90 to 99 percent.
Anthropic may get a 500 megawatt data center in Osaka, showing the scale of compute deals.
Zuckerberg and Wang interview on Meta's Muse app, Muse Wonder and agentic commerce.
Omron buys OnLogic to push into physical AI hardware.
Show 7 more
Chinese mathematicians form an alliance to build math-focused AI models.
Banks limit exposure on a large AI startup loan, a sign of caution in AI financing.
Anthropic warns small models may soon self-replicate on cheap IoT devices like Android TV boxes.
Epic reportedly tells developers to stop hand-writing code, a bold AI-first production bet.
Suki raises 90 million dollars for its AI medical scribe.
Bumble is in talks to go private.
Meta's Muse Wonder lands near the top of a major intelligence benchmark.
About this issue
Source health
Which sources were fetched today, and how many items each returned.
| Source | Section | Items | Status |
|---|---|---|---|
| GitHub Trending | Main | 9 | ok |
| Product Hunt | Main | 50 | ok |
| Show HN | Main | 7 | ok |
| HN front page | Main | 30 | ok |
| YC Launches | Main | 20 | ok |
| HF trending models | Main | 20 | ok |
| HF trending spaces | Main | 20 | ok |
| HF daily papers | Main | 20 | ok |
| Lobsters | Main | 25 | ok |
| Dev.to top | Main | 20 | ok |
| Techmeme | Main | 15 | ok |
| AndroidX releases | Android · Official & releases | 37 | ok |
| Android Developers Blog | Android · Official & releases | 4 | ok |
| Android Studio releases | Android · Official & releases | 3 | ok |
| Kotlin blog | Android · Official & releases | 4 | ok |
| Kotlin releases | Android · Official & releases | 0 | ok |
| Compose Multiplatform releases | Android · Official & releases | 1 | ok |
| Android Developers YouTube | Android · Official & releases | 0 | failed |
| Kotlin on YouTube (JetBrains) | Android · Official & releases | 0 | failed |
| Google Developers Blog | Android · Official & releases | 2 | ok |
| AndroidX merged API changes | Android · AOSP / AndroidX insider | 9 | ok |
| Dove Letter (skydoves) | Android · Experts | 5 | ok |
| skydoves on Medium | Android · Experts | 0 | ok |
| Jake Wharton | Android · Experts | 0 | ok |
| Zac Sweers | Android · Experts | 0 | ok |
| Chris Banes | Android · Experts | 2 | ok |
| Ryan Harter | Android · Experts | 0 | ok |
| Talking Kotlin | Android · Experts | 0 | ok |
| Joe Birch | Android · Experts | 2 | ok |
| CommonsWare (Mark Murphy) | Android · Experts | 2 | ok |
| TechYourChance (Vasiliy Zukanov) | Android · Experts | 1 | ok |
| Kt. Academy (Marcin Moskała) | Android · Experts | 2 | ok |
| Touchlab (Kotlin Multiplatform) | Android · Experts | 3 | ok |
| Landscapist releases | Android · Library releases | 0 | ok |
| Balloon releases | Android · Library releases | 0 | ok |
| Sandwich releases | Android · Library releases | 0 | ok |
| Cloudy releases | Android · Library releases | 0 | ok |
| Compose Stability Analyzer releases | Android · Library releases | 0 | ok |
| Haze releases | Android · Library releases | 1 | ok |
| Compose Unstyled releases | Android · Library releases | 0 | ok |
| Coil releases | Android · Library releases | 0 | ok |
| Circuit releases | Android · Library releases | 0 | ok |
| Metro releases | Android · Library releases | 0 | ok |
| Molecule releases | Android · Library releases | 0 | ok |
| Turbine releases | Android · Library releases | 0 | ok |
| Paparazzi releases | Android · Library releases | 0 | ok |
| SQLDelight releases | Android · Library releases | 1 | ok |
| OkHttp releases | Android · Library releases | 0 | ok |
| Retrofit releases | Android · Library releases | 0 | ok |
| LeakCanary releases | Android · Library releases | 0 | ok |
| Koin releases | Android · Library releases | 0 | ok |
| Dagger / Hilt releases | Android · Library releases | 0 | ok |
| KSP releases | Android · Library releases | 0 | ok |
| kotlinx.coroutines releases | Android · Library releases | 0 | ok |
| kotlinx.serialization releases | Android · Library releases | 0 | ok |
| Ktor releases | Android · Library releases | 0 | ok |
| Apollo Kotlin releases | Android · Library releases | 0 | ok |
| Kermit releases | Android · Library releases | 0 | ok |
| Store releases | Android · Library releases | 0 | ok |
| Decompose releases | Android · Library releases | 0 | ok |
| Orbit MVI releases | Android · Library releases | 0 | ok |
| Roborazzi releases | Android · Library releases | 1 | ok |
| detekt releases | Android · Library releases | 0 | ok |
| ktlint releases | Android · Library releases | 0 | ok |
| MockK releases | Android · Library releases | 0 | ok |
| Lottie releases | Android · Library releases | 0 | ok |
| Authority Insights (leaks & teardowns) | Android · Leaks & ecosystem | 8 | ok |
| 9to5Google · Android | Android · Leaks & ecosystem | 8 | ok |
| Android Authority | Android · Leaks & ecosystem | 6 | ok |
| Android Faithful | Android · Leaks & ecosystem | 1 | ok |
| Android Weekly | Android · Community | 15 | ok |
| GitHub Trending · Kotlin | Android · Community | 17 | ok |
| Lobsters · android | Android · Community | 2 | ok |
| HN · Android | Android · Community | 7 | ok |
| ProAndroidDev | Android · Community | 5 | ok |
| Kotlin Weekly | Android · Community | 1 | ok |
| Philipp Lackner (YouTube) | Android · Community | 0 | failed |