SHUBHAM SINGH BISHT

Android Daily ·

Android's official developer blog moves to developer.android.com, with a new feed

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 · 3 min listen
  1. #1Official

    The official Android Developers Blog moves to a new home

    Why it matters If you follow Android news by feed reader, switch to the new feed or you may miss official announcements.

    Google closed the Android Developers Blog on Blogger and moved it to the blog section of the main Android developer site. The new blog has better formatting and sits next to the technical docs. Google asks readers to update bookmarks and feed readers today. The post lists a new feed, and it does not say whether the old one will keep updating.

    Read the original · Android Developers Blog

  2. #2Official

    Jetpack Compose 1.13 reaches beta

    Why it matters Compose 1.13 is close to stable, so now is the time to test it and report bugs.

    Version 1.13 of Jetpack Compose, Android's UI toolkit, is now in beta. This line makes the FlexBox and Grid layouts stable, adds polygon shapes, and deprecates the older text field that takes a plain string. The beta fixes a bug where the keyboard did not appear if a text field asked for focus too early. It also needs Kotlin 2.2.

    Read the original · AndroidX releases

  3. #3Official

    AndroidX Collection 1.7.0 is stable

    Why it matters A safe update that adds sorting and searching to the fast collections Compose already uses.

    AndroidX Collection, the library of memory-efficient lists, sets and maps used inside Compose, released version 1.7.0 as stable. It adds sorting to object lists, binary search to object and number lists, and a faster get-or-put on its hash map. Float keys now hash with fewer collisions. The release notes list no breaking changes.

    Read the original · AndroidX releases

  4. #4Official

    Compose Multiplatform 1.12.2 fixes an iOS text field crash

    Why it matters If you ship a Compose Multiplatform app on iOS, this patch removes a crash and some visual glitches.

    JetBrains released Compose Multiplatform 1.12.2, a bug fix update for its toolkit that shares Compose UI across Android, iOS, desktop and web. On iOS, it fixes a crash when a text field lost focus because a dialog opened. It also fixes flickering shadows on animated shapes, and adapts text input focus on the web for Safari 27 on iOS.

    Read the original · Compose Multiplatform releases

  5. #5Ecosystem

    Samsung brings Android 17 to the Galaxy S23 and more phones

    Why it matters Android 17 is reaching millions more Samsung phones, so check how your app behaves on it now.

    Samsung's One UI 9 update, based on Android 17, reached the Galaxy S23 and several more Galaxy devices, according to 9to5Google. Samsung ships to many of the world's Android users, so each new device brings a large group onto the new version. Meanwhile Sony canceled a promised Android 17 update for one of its Xperia phones.

    Read the original · 9to5Google · Android

02 · Reference

Go deeper: every Android item, by lane

Official & releases

52The official Android blog now lives on the main developer site with a new feed, Compose 1.13 is in beta and needs Kotlin 2.2, Collection 1.7.0 is a safe stable update, and Compose Multiplatform 1.12.2 fixes an iOS text field crash.

AndroidX library releases

Stable · safe to ship · 5

Beta · API frozen, try it before it ships · 15

Alpha · experimental, APIs may change · 15
Show 9 more

AOSP / AndroidX insider

32AndroidX changes point to Glance tiles becoming widgets on Wear OS, Perfetto counter metrics in Macrobenchmark, more interactive Remote Compose, and Navigation3 state APIs still in flux.
Show 24 more

Experts

17Read Doveletter's new series on how the Compose compiler turns a function into a composable, and Chris Banes on Haze 2.0 and Compose's progressive blur.
Show 9 more

Library releases

3SQLDelight 2.4.1 and Compose Unstyled 2.10.1 are safe patch updates, while Decompose 3.6.0-beta02 adds Linux targets and fixes predictive back in right-to-left layouts.

Leaks & ecosystem

22Samsung's Android 17 update reaches the Galaxy S23 while Sony cancels one Xperia update, Google is retiring the Heads Up safety tool, and Gemini subscribers lose Pro model access in mid-November.
Show 14 more

Community

32See jadx-headless-mcp for agent-driven APK analysis, Google's android/skills repository for coding agents, and Tilt's two-part story on tracing a Kotlin Multiplatform data layer.
Show 24 more
Start the Product & AI read4 stories, ranked by what you’d regret missing · 2 min read · 2 min listen
  1. #1Across sources

    Meta releases Muse Spark Open, a very large open model for coding

    Why it matters Teams that host their own models now have a frontier-scale option for coding agents, free under Meta's license.

    Meta released Muse Spark Open, an open-weight model with 816 billion parameters, of which 42 billion work on each token. It reads text and images, handles a context of one million tokens, and is built for coding and agent tasks. Weights are on Hugging Face and three clouds, and seven providers offer it through an API. Meta has not yet published capability scores.

    Read the original · HN front pageDiscussionProduct notes

  2. #2Fastest rising

    rea lets coding agents reverse engineer apps and binaries

    Why it matters Agents can now take apart Android apps, binaries and firmware with the same tools experts use.

    rea is an open-source tool that topped GitHub trending, gaining about 32 thousand stars in a day. It is an MCP server, a plug-in that coding agents such as Claude Code or Codex can call. It gives them reverse engineering tools like Ghidra for binaries and JADX and Apktool for Android apps. Analysis runs on your own machine.

    Read the original · GitHub TrendingProduct notes

  3. #3New idea

    Perplexity gives away the sandbox it uses to contain AI agents

    Why it matters Containing agents is becoming shared infrastructure that teams can adopt instead of building.

    Perplexity open-sourced SPACE, the sandbox it uses to stop AI agents from going rogue, Semafor reports. A sandbox is a sealed environment where an agent can run code without touching anything else. Cloud companies CoreWeave, Akamai and Crusoe agreed to offer it to their customers. On the same day, Agent37 launched low-cost agent sandboxes on Hacker News.

    Read the original · TechmemeProduct notes

  4. #4In 2 sources

    Unfenced finds land for sale next to public land

    Why it matters Joining two free public datasets produced a product that topped Show HN and the front page.

    Unfenced is a search site for land that borders public land, such as national forests or wilderness. It checks listings against official boundary maps from the US Geological Survey. From nearly 700 thousand active listings, it found about 13,600 qualifying properties. A small design studio built it, and it led both Show HN and the Hacker News front page.

    Read the original · Show HNDiscussionProduct notes

02 · Reference

Go deeper: themes, product notes, every source

Themes across sources

6Agent sandboxes · AI-driven reverse engineering · Open-weight frontier models · On-device AI · Small decision models · Agent harnesses

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

Agent sandboxes

steady5 items3 sources

Containing and supervising agents is its own market. Perplexity open-sourced its agent sandbox with cloud partners, Agent37 launched cheap sandboxes on Hacker News, Buddy added isolated machines for coding agents, Opsin raised money to watch agents at work, and withHuman adds a human approval step.

Why now. Agents now run long tasks unattended with real permissions, so isolation, cost and approval have become operational needs.

Perplexity open sources SPACE, its AI agent sandbox, and says partners including CoreWeave, Akamai, and Crusoe agreed to offer SPACE to customers (Reed Albergotti/Semafor)Techmeme

AI-driven reverse engineering

steady4 items3 sources

Agents are being handed reverse engineering tools. rea topped GitHub with one MCP server for binaries, APKs and firmware, a front-page post used AI to reverse engineer hardware, a headless JADX server lets agents read Android APKs, and Ghidra itself is trending again.

Why now. MCP gives agents a standard way to drive command-line analysis tools, and models are now good at reading decompiled code.

morluto/reaGitHub Trending

Open-weight frontier models

new5 items2 sources

Big open-weight releases keep landing. Meta's Muse Spark Open led Hacker News and trends on Hugging Face, alongside Qwen3.8 27B, MiniCPM-V 4.7 and Liquid AI's LFM2.5.

Why now. Mixture-of-experts models keep the active size small, so very large open models are practical to host and to offer through many API providers.

Muse Spark OpenHN front page

More

On-device AI

↑ rising6 items4 sources

Small models keep moving work onto your own hardware. EmbeddingGemma 2 is the top model on Hugging Face, a guide sizes it for small servers, MiniCPM-V ships a 1 billion parameter version, and several launches convert documents or bank statements without uploading them.

Why now. Small embedding and vision models now perform well enough on laptops and phones, and privacy is a selling point.

google/embeddinggemma-2HF trending models

Small decision models

steady6 items4 sources

Models that only make quick, typed choices keep spreading. A front-page post tests the Jev decision model on trolley-problem dilemmas, a decision-model leaderboard trends on Hugging Face, Nimble and Brussle launched on Product Hunt, and two new papers train decision models that share control with language models.

Why now. Agents make many small routing and approval calls per task, so cheap, fast, typed decisions matter more than raw intelligence.

Jev meets the Trolley ProblemHN front page

Agent harnesses

steady6 items4 sources

The programs that run coding agents are now products. Superpowers, ECC and ruflo keep climbing GitHub, OpenMuse launched a self-hosted assistant for any harness, a Hugging Face space trains across several harnesses, and the Unified Harness Protocol aims to standardize them.

Why now. Model quality has converged enough that the workflow, memory and rules around the model decide results.

obra/superpowersGitHub Trending

More

Fading: Agent-ready Android (last seen 2026-10-15) · Agent memory layers (last seen 2026-10-15) · Verifying agent work (last seen 2026-10-15) · Coding-agent control surfaces (last seen 2026-10-13) · MCP-native apps (last seen 2026-10-13) · Self-improving agents (last seen 2026-10-12)

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

discussionHN front pagehype flag

Muse Spark Open

676 pts · 365 comments

What
Meta released Muse Spark Open, an open-weight model with 816 billion parameters, 42 billion active per token, a one million token context, and image and text input, built for coding and agent work.
For whom
Teams that want a frontier-class coding and agent model they can host themselves or rent from many API providers.
Problem
The strongest coding models are closed, so teams that need control over hosting, cost or data have to settle for weaker open ones.
Why now
Mixture-of-experts designs keep the active size near 42 billion, which makes an 816 billion model practical to serve; Meta says it even runs on two Mac Studios.
Wedge
Day-one availability on Hugging Face, three clouds and seven API providers, so trying it costs no setup.
Money
Free under the Muse Model License with limited commercial terms; Meta does not sell it directly.
Like
Qwen3.8 · DeepSeek V4 · Kimi K3 · Llama 4

The launch page has a benchmark heading but no capability scores, and the full evaluation report is still to come.

LessonLaunch where developers already are: many hosts on day one turned an open-weight release into the top story on Hacker News.

repoGitHub Trending

morluto/rea

★ 32.3k today

What
rea is an open-source MCP server that lets coding agents reverse engineer software, from native binaries and Android APKs to Electron apps, websites and firmware.
For whom
Security researchers and engineers who need to understand software they did not write, using an agent such as Claude Code, Codex, Cursor or Gemini CLI.
Problem
Reverse engineering means juggling many tools, such as Ghidra, JADX, Apktool and tshark, and stitching their output together by hand.
Why now
Agents can now drive command-line tools reliably through MCP, so one server can hand them a full reverse engineering toolbox.
Wedge
One setup command registers it with the popular coding agents and reuses tools you already have installed.
Money
MIT-licensed open source, no business model yet; commercial tools such as IDA must be licensed separately.
Like
Ghidra · IDA Pro · GhidraMCP · jadx-mcp

LessonWrap an expert workflow, not a single tool: the value is that the agent gets every reverse engineering tool behind one interface.

newsTechmeme

Perplexity open sources SPACE, its AI agent sandbox, and says partners including CoreWeave, Akamai, and Crusoe agreed to offer SPACE to customers (Reed Albergotti/Semafor)

What
Perplexity open-sourced SPACE, the sandbox it uses to contain AI agents, and says CoreWeave, Akamai and Crusoe will offer it to customers.
For whom
Companies running AI agents that execute code or browse, which need to stop a misbehaving agent from doing damage.
Problem
Agents with tool access can delete data, leak secrets or run away with compute, and most teams build their own containment.
Why now
Agents now run unattended for long tasks, so containment has moved from an option to a requirement.
Wedge
Giving away a sandbox that is already proven in production, with infrastructure partners distributing it.
Money
Open source, no model stated; it likely builds goodwill and an ecosystem around Perplexity's agent stack.
Like
E2B · Daytona · Modal Sandboxes · Firecracker

LessonOpen-sourcing internal infrastructure can turn a cost center into distribution when partners resell it.

productShow HNHN front page

Show HN: Unfenced – Land for sale that borders public land

277 pts · 186 comments

What
Unfenced lists land for sale that touches public land, such as national forests or wilderness, by checking listings against official boundary maps.
For whom
People shopping for rural land who want it to border public land for hunting, hiking or privacy.
Problem
Listings rarely say whether a parcel touches public land, so buyers check maps by hand one property at a time.
Why now
Public boundary data such as the US Geological Survey's protected areas database is free and detailed enough to match against parcel data at scale.
Wedge
One narrow question answered for every listing: about 13,600 qualifying properties found among nearly 700 thousand searched.
Money
No pricing; it is a side project of a design studio with a tip link.
Like
LandWatch · onX Hunt · Zillow

LessonJoin two public datasets to answer a question listings never answer; the narrow filter is the product.

modelHF trending models

google/embeddinggemma-2

trending 999

What
EmbeddingGemma 2 is Google DeepMind's small embedding model: about 740 million parameters that turn text, code, images, video and audio into one shared vector space.
For whom
App developers building search, retrieval or classification that should run on a phone or laptop.
Problem
Semantic search over mixed media usually needs a cloud embedding API, which adds cost, delay and privacy risk.
Why now
Small models now match larger ones on retrieval, and phones have enough memory to run them.
Wedge
One small model for every media type, with outputs you can shrink from 768 to 128 numbers to save space.
Money
Apache 2.0 open weights; it pulls developers toward the Gemma family and Google's tooling.
Like
OpenAI text-embedding-3 · Nomic Embed · Jina embeddings · EmbeddingGemma

LessonShrinkable outputs let one model fit both a phone and a server; design for the smallest target first.

discussionHN · AndroidAndroid

Launch HN: Agent37 (YC F26) – Low-cost agent sandboxes for running agents

160 pts · 124 comments

What
Agent37 sells sandboxes for running AI agents, at what it says is under a tenth of the cost of most sandbox providers.
For whom
Teams that run many agents and pay per sandbox; it says customers spend about 37 cents a month per sandbox.
Problem
Sandbox providers price for heavy compute, which is costly when most agents sit idle.
Why now
Agent workloads have grown from demos to fleets, so per-sandbox cost now shows up on the bill.
Wedge
Price: a launch built around an order-of-magnitude cheaper sandbox.
Money
Usage-based sandbox hosting.
Like
E2B · Daytona · Modal

LessonWhen a category is new, a large price cut can be the whole wedge; it works only if your costs are structurally lower.

productProduct Hunt

AI Sandboxes by Buddy

#30 on Product Hunt

What
Buddy, a CI and deployment company, now offers isolated cloud machines for AI coding agents.
For whom
Teams that let coding agents run builds and tests and want them off developer laptops.
Problem
Agents running locally can damage a machine or leak credentials, and parallel agents compete for one computer.
Why now
Coding agents now run long tasks in parallel, which needs disposable machines.
Wedge
Reusing the build infrastructure Buddy already runs for CI.
Money
Paid cloud machines, likely metered like its CI minutes.
Like
E2B · GitHub Codespaces · Daytona

LessonAdjacent infrastructure is a fast route into a new category: a CI company already knows how to run throwaway machines.

productProduct Hunt

withHuman

#11 on Product Hunt

What
withHuman is an open-source layer that pauses an AI agent and asks a person to approve an action before it runs.
For whom
Teams deploying agents that touch money, customers or production systems.
Problem
Teams either trust agents fully or hard-code approval steps into each workflow.
Why now
Agents are now trusted with real actions, and regulators and customers ask who approved them.
Wedge
Open source and narrow: just the approval step, which any agent framework can call.
Money
Open source, no model stated; a hosted version with audit logs is the likely path.
Like
HumanLayer · LangGraph interrupts · Slack approval bots

LessonSell the missing control, not the agent: a small trust layer can ride every agent framework.

productProduct Hunthype flag

Tiden

#42 on Product Hunt

What
Tiden checks whether a coding agent's changes match what you asked it to do.
For whom
Developers who review pull requests written by coding agents.
Problem
Agents can drift from the request, and reviewers have to read every diff to notice.
Why now
Agents now open many pull requests a day, more than people can review closely.
Wedge
A single question, did the change match the intent, rather than a full code review.
Money
Not stated.
Like
CodeRabbit · Graphite Reviewer · Polylane

The listing gives a one-line claim with no detail on how intent is checked or how accurate it is.

LessonVerification grows with generation: the more code agents write, the more a narrow checker is worth.

repoGitHub Trending · KotlinAndroid

1013503897/jadx-headless-mcp

★ 9 today

What
jadx-headless-mcp is an MCP server for static analysis of Android APKs, built directly on the JADX decompiler with no desktop app needed.
For whom
Android security researchers and developers who want an agent to inspect an APK.
Problem
JADX is a desktop tool, so an agent cannot drive it without a person clicking through the interface.
Why now
Coding agents speak MCP, so a headless wrapper turns a classic tool into agent-ready infrastructure.
Wedge
Headless and agent-first: it runs on a server or in CI.
Money
Open source, no model.
Like
JADX · rea · apktool

LessonMany classic developer tools are one headless wrapper away from being agent tools.

repoGitHub Trending · KotlinAndroid

android/skills

★ 35 today

What
android/skills is a repository of Agent Skills for Android development in the official Android GitHub organization, trending in Kotlin today.
For whom
Android developers using coding agents who want them to follow current Android practice.
Problem
Agents often write outdated Android code because their training data is older than the current APIs.
Why now
Google is putting agents at the center of Android tooling, including the Android CLI, which now ships Android skills.
Wedge
Skills are plain files any agent can load, so they work outside Google's own tools.
Money
Free; it supports Google's Android and Gemini tools.
Like
Android CLI · Gemini in Android Studio · Cursor rules

LessonPlatform owners can steer AI-written code by publishing guidance in the format agents read.

productYC Launches

PlaceCall by VOYGR – The phone is the last API

67 votes

What
PlaceCall by VOYGR is an API that finds businesses and calls them in parallel to get answers, quotes and bookings.
For whom
Apps and agents that need information only available by phone, such as prices, availability or bookings.
Problem
Many local businesses have no API or online booking, so getting an answer means a person on the phone.
Why now
Voice models can now hold short business calls, which makes the phone callable from code.
Wedge
Calling many businesses at once and returning structured answers.
Money
Usage-based API pricing.
Like
Bland AI · Vapi · Google Duplex

LessonTreat the oldest interface as an API: where there is no integration, an agent can be the integration.

productShow HN

Show HN: Sygal – Convert PDFs, scans and web pages to Markdown, locally

34 pts · 11 comments

What
Sygal converts PDFs, scanned documents and web pages to Markdown on your own computer.
For whom
People feeding documents to AI tools who cannot upload them to a cloud service.
Problem
Cloud converters see your documents, and local tools handle scans and layouts poorly.
Why now
Small local vision and OCR models are now good enough to read scans on a laptop.
Wedge
Local processing as the main promise.
Money
Not stated.
Like
Marker · Docling · MinerU

LessonPrivacy is a feature you can lead with when the input is a document people would not email.

productShow HN

Show HN: Flipflop – Iroh-powered persistent tunnels for filesharing

117 pts · 62 comments

What
Flipflop is an open-source app that keeps persistent, direct tunnels between devices for sharing files, built on the Iroh networking library.
For whom
People moving files between their own devices or with trusted people, who found LocalSend-style tools limiting.
Problem
Local transfer apps only work on the same network, and cloud sharing goes through a third party.
Why now
Peer-to-peer libraries such as Iroh handle connection setup across networks, so an app can skip running a relay service.
Wedge
A standing connection to trusted contacts instead of one-off transfers.
Money
Open source, no model.
Like
LocalSend · Syncthing · Tailscale Taildrop

LessonA new networking layer makes old product ideas cheap again: check what infrastructure now comes free.

repoGitHub Trendinghype flag

mksglu/context-mode

★ 904 today

What
context-mode shrinks what tool output takes up in a coding agent's context, saves session memory, and routes calls across many agent platforms through MCP and hooks.
For whom
Developers whose coding agents burn context on long command output.
Problem
Raw tool output fills the context window, which raises cost and makes the agent forget earlier work.
Why now
Long agent sessions are now common, so context is the scarce resource.
Wedge
It claims a 98 percent reduction in tool output, a single measurable promise.
Money
Open source, no model.
Like
caveman · xerj · Claude Code compaction

The 98 percent figure is the project's own claim, with no independent test.

LessonWhen a resource gets scarce, a tool that saves it sells itself; lead with the number.

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

13
  • morluto/reacard

    Reverse engineer anything with agents, from app behavior down to native binaries.

    ★ 32.3k today

  • anthropics/claude-plugins-officialnew

    Anthropic now curates an official directory of Claude Code plugins, a sign plugins are becoming a distribution channel.

    ★ 307 today

  • K-Dense-AI/scientific-agent-skills

    177 ready-made skills that turn general agents into science assistants; skills as a distribution format.

    ★ 1,091 today

  • JuliusBrussee/cavemannew

    A viral joke skill with a real point: terse agent output can cut token costs sharply.

    ★ 433 today

  • debpalash/VoiceStudio

    Voice cloning, dubbing and transcription that run fully on your own machine, as an open-source ElevenLabs alternative.

    ★ 3,024 today

  • obra/superpowersnew

    A widely used skills framework that gives coding agents a repeatable development method.

    ★ 582 today

  • ruvnet/ruflonew

    A harness for coordinating swarms of agents, still climbing as multi-agent setups spread.

    ★ 147 today

  • mksglu/context-modecard

    Context window optimization for AI coding agents. Sandboxes tool output (98% reduction), persists session memory, and enforces routing across 17 platforms via MCP + hooks.

    ★ 904 today

Show 5 more
  • NationalSecurityAgency/ghidra

    The NSA's reverse engineering suite trends alongside agent-driven reverse engineering tools.

    ★ 1,369 today

  • affaan-m/ECCnew

    Packages skills, memory and security rules for several coding agents at once; the harness is becoming the product.

    ★ 1,013 today

  • lightseekorg/tokenspeednew

    A new LLM inference engine chasing raw speed; worth watching if you serve open models.

    ★ 30 today

  • xerj-org/xerjnew

    Indexes code and data so agents stop burning tokens on repeated grep and sed searches.

    ★ 715 today

  • LibreSpark/LibreTVnew

    A self-hosted video site builder trending on GitHub; mostly of interest to Chinese-speaking users.

    ★ 978 today

Product Hunt

50
  • GLM-Legalnew

    Z.ai ships a model tuned for contract review, part of the move to vertical legal models.

    #1 on Product Hunt

  • OpenMusenew

    A self-hostable personal assistant that plugs into whichever agent harness you already use.

    #2 on Product Hunt

  • Nostaliganew

    Turns your listening history into a memory journal; a nostalgia angle on music data.

    #3 on Product Hunt

  • Atticusnew

    Push-to-talk control of your Mac: hold a key, say what you want, and it acts.

    #4 on Product Hunt

  • Audimatenew

    An assistant that works inside existing apps instead of asking you to switch to a chat window.

    #5 on Product Hunt

  • SaidTracenew

    Private voice memos where you pick which AI model processes them.

    #6 on Product Hunt

  • Navigenew

    Budget caps and cost controls for AI agents before a runaway loop drains your account.

    #7 on Product Hunt

  • Lleveragenew

    Agents for back-office work in logistics and manufacturing, where paperwork still runs on email.

    #8 on Product Hunt

Show 42 more
  • CinemaLutnew

    Extracts the color grade of any film frame as a reusable lookup table.

    #9 on Product Hunt

  • Chatway MCPnew

    Lets ChatGPT or Claude answer support tickets through MCP, with no custom integration.

    #10 on Product Hunt

  • withHumannewcard

    A open-source human approval layer for AI Agents

    #11 on Product Hunt

  • siliXonnew

    Describe a circuit in plain English and get a board you can order; hardware design meets text prompts.

    #12 on Product Hunt

  • SuperLocalMemorynew

    Agent memory that runs locally with small models such as Gemma.

    #13 on Product Hunt

  • Talanoanew

    Email redesigned to feel like a chat thread.

    #14 on Product Hunt

  • Backdropnew

    A builder for AI coworkers that handle a team's repeated workflows.

    #15 on Product Hunt

  • Dreamscale Labsnew

    Robot AI models served from the cloud, so robots do not need large onboard compute.

    #16 on Product Hunt

  • Reelbirdnew

    An open-source screen recorder that combines Loom-style sharing with polished Screen Studio-style edits.

    #17 on Product Hunt

  • Kerno 1.0new

    An AI agent that tests backend behavior at runtime, aimed at QA for APIs.

    #18 on Product Hunt

  • Ubuntu 26.10 (Stonking Stingray)new

    Ubuntu 26.10 ships GNOME 51 and Rust-based core utilities, a notable step for Rust in Linux basics.

    #19 on Product Hunt

  • Daddy Issuesnew

    A weekly family planning briefing aimed at fathers.

    #20 on Product Hunt

  • Polylane Impact Analysisnew

    Flags pull requests likely to break production before they merge.

    #21 on Product Hunt

  • Brusslenew

    Tunes the Jev decision model on your own data to improve its accuracy.

    #22 on Product Hunt

  • Superdraftnew

    Finds moments to email customers based on their activity, then drafts the email.

    #23 on Product Hunt

  • Rushilonew

    Edit video by giving instructions to Claude Code or Codex; coding agents as a general command line.

    #24 on Product Hunt

  • Apple HomeViewnew

    Apple's new home dashboard for connected devices.

    #25 on Product Hunt

  • Basedash Connector gallerynew

    Shows what you can build from a data source before you connect it, which lowers the first step.

    #26 on Product Hunt

  • Afterpartynew

    Turns meeting outcomes into follow-up actions automatically.

    #27 on Product Hunt

  • Shrinenew

    A place to keep, run and manage your AI agents.

    #28 on Product Hunt

  • Fwrdnew

    Forwards iPhone texts into Slack, Discord or Telegram.

    #29 on Product Hunt

  • AI Sandboxes by Buddynewcard

    Isolated cloud machines for your AI coding agents

    #30 on Product Hunt

  • Audio Lip Sync Pro for Blendernew

    A Blender add-on that animates faces from an audio track.

    #31 on Product Hunt

  • CmdRoundnew

    Tests how engineering candidates work with AI tools, a new kind of technical interview.

    #32 on Product Hunt

  • NestJS Observenew

    Observability for NestJS apps with no setup.

    #33 on Product Hunt

  • Actually Askednew

    A branded FAQ page that updates itself from the questions people really ask.

    #34 on Product Hunt

  • CodeCrabnew

    Another AI code reviewer, competing on analysis depth and integrations.

    #35 on Product Hunt

  • Confidence by Spotify

    Spotify offers its product experimentation tooling to other teams, pitched as speed with control.

    #36 on Product Hunt

  • PixLate

    Translates documents while keeping artwork and layout.

    #37 on Product Hunt

  • Screenwell.ai

    An AI interviewer for first-round job screening.

    #38 on Product Hunt

  • AirJelly

    A second brain that organises itself and makes suggestions; memory moving from storage to action.

    #39 on Product Hunt

  • daily.dev for agents

    A discovery feed for agents rather than people; content platforms court agent readers.

    #40 on Product Hunt

  • Payload Peek

    Catches webhooks and generates TypeScript types from their payloads.

    #41 on Product Hunt

  • Tidencard

    Intent verification for coding agents

    #42 on Product Hunt

  • Inferock

    Serverless inference that refunds credits for failed calls; pricing on reliability.

    #43 on Product Hunt

  • Orbora

    Generates personalised guided audio such as meditation and breathing.

    #44 on Product Hunt

  • SupoClip

    Open-source clipper that turns long videos into short ones.

    #45 on Product Hunt

  • Winq

    Controls a Mac with eye gaze and winks.

    #46 on Product Hunt

  • Nimble

    Another decision model launch, with a state-of-the-art claim and no detail in the listing.

    #47 on Product Hunt

  • Microsoft Maia 300

    Microsoft's in-house AI chip, pitched at multi-gigawatt scale; hyperscalers keep building their own silicon.

    #48 on Product Hunt

  • Rambox

    Workspaces that switch with your task context.

    #49 on Product Hunt

  • Pancake

    Pitched as Vercel for go-to-market: deploy-style workflows for sales and marketing.

    #50 on Product Hunt

Show HN

8

HN front page

30
Show 22 more

YC Launches

20
Show 12 more

HF trending models

18
Show 10 more

HF trending spaces

15
Show 7 more

HF daily papers

20
Show 12 more

Lobsters

25
Show 17 more

Dev.to top

20
Show 12 more

Techmeme

15
Show 7 more

About this issue

388 items77 sources5 seen in several sources4 sources failed

Source health

77Which sources were fetched today, and how many items each returned.
SourceSectionItemsStatus
GitHub TrendingMain13ok
Product HuntMain50ok
Show HNMain8ok
HN front pageMain30ok
YC LaunchesMain20ok
HF trending modelsMain20ok
HF trending spacesMain20ok
HF daily papersMain20ok
LobstersMain25ok
Dev.to topMain20ok
TechmemeMain15ok
AndroidX releasesAndroid · Official & releases37ok
Android Developers BlogAndroid · Official & releases5ok
Android Studio releasesAndroid · Official & releases2ok
Kotlin blogAndroid · Official & releases5ok
Kotlin releasesAndroid · Official & releases0ok
Compose Multiplatform releasesAndroid · Official & releases4ok
Android Developers YouTubeAndroid · Official & releases0failed
Kotlin on YouTube (JetBrains)Android · Official & releases0failed
Google Developers BlogAndroid · Official & releases1ok
AndroidX merged API changesAndroid · AOSP / AndroidX insider32ok
Dove Letter (skydoves)Android · Experts5ok
skydoves on MediumAndroid · Experts0ok
Jake WhartonAndroid · Experts0ok
Zac SweersAndroid · Experts0ok
Chris BanesAndroid · Experts2ok
Ryan HarterAndroid · Experts0ok
Talking KotlinAndroid · Experts0ok
Joe BirchAndroid · Experts2ok
CommonsWare (Mark Murphy)Android · Experts2ok
TechYourChance (Vasiliy Zukanov)Android · Experts1ok
Kt. Academy (Marcin Moskała)Android · Experts2ok
Touchlab (Kotlin Multiplatform)Android · Experts3ok
Landscapist releasesAndroid · Library releases0ok
Balloon releasesAndroid · Library releases0ok
Sandwich releasesAndroid · Library releases0ok
Cloudy releasesAndroid · Library releases0ok
Compose Stability Analyzer releasesAndroid · Library releases0ok
Haze releasesAndroid · Library releases0ok
Compose Unstyled releasesAndroid · Library releases1ok
Coil releasesAndroid · Library releases0ok
Circuit releasesAndroid · Library releases0ok
Metro releasesAndroid · Library releases0ok
Molecule releasesAndroid · Library releases0ok
Turbine releasesAndroid · Library releases0ok
Paparazzi releasesAndroid · Library releases0ok
SQLDelight releasesAndroid · Library releases1ok
OkHttp releasesAndroid · Library releases0ok
Retrofit releasesAndroid · Library releases0ok
LeakCanary releasesAndroid · Library releases0ok
Koin releasesAndroid · Library releases0ok
Dagger / Hilt releasesAndroid · Library releases0ok
KSP releasesAndroid · Library releases0ok
kotlinx.coroutines releasesAndroid · Library releases0ok
kotlinx.serialization releasesAndroid · Library releases0ok
Ktor releasesAndroid · Library releases0ok
Apollo Kotlin releasesAndroid · Library releases0ok
Kermit releasesAndroid · Library releases0ok
Store releasesAndroid · Library releases0ok
Decompose releasesAndroid · Library releases1ok
Orbit MVI releasesAndroid · Library releases0ok
Roborazzi releasesAndroid · Library releases0ok
detekt releasesAndroid · Library releases0ok
ktlint releasesAndroid · Library releases0ok
MockK releasesAndroid · Library releases0ok
Lottie releasesAndroid · Library releases0ok
Authority Insights (leaks & teardowns)Android · Leaks & ecosystem8ok
9to5Google · AndroidAndroid · Leaks & ecosystem8ok
Android AuthorityAndroid · Leaks & ecosystem6ok
Android FaithfulAndroid · Leaks & ecosystem1ok
Android WeeklyAndroid · Community0failed
GitHub Trending · KotlinAndroid · Community13ok
Lobsters · androidAndroid · Community3ok
HN · AndroidAndroid · Community10ok
ProAndroidDevAndroid · Community5ok
Kotlin WeeklyAndroid · Community1ok
Philipp Lackner (YouTube)Android · Community0failed
1 of 5
#1Official

The official Android Developers Blog moves to a new home

Why it matters If you follow Android news by feed reader, switch to the new feed or you may miss official announcements.

Google closed the Android Developers Blog on Blogger and moved it to the blog section of the main Android developer site. The new blog has better formatting and sits next to the technical docs. Google asks readers to update bookmarks and feed readers today. The post lists a new feed, and it does not say whether the old one will keep updating.

Read the original · Android Developers Blog

1 of 4
#1Across sources

Meta releases Muse Spark Open, a very large open model for coding

Why it matters Teams that host their own models now have a frontier-scale option for coding agents, free under Meta's license.

Meta released Muse Spark Open, an open-weight model with 816 billion parameters, of which 42 billion work on each token. It reads text and images, handles a context of one million tokens, and is built for coding and agent tasks. Weights are on Hugging Face and three clouds, and seven providers offer it through an API. Meta has not yet published capability scores.

Read the original · HN front pageDiscussion