We beat Google DeepMind but got killed by Zhipu AI
Two months ago, my friends in AI and I asked: What if an AI could actually use a phone like a human? So we built an agentic framework that taps, swipes, types… and somehow it’s outperforming giant labs like Google DeepMind and Microsoft Research on the AndroidWorld benchmark. We were thrilled about our results until a massive lab (Zhipu AI) released its results last week to take the top spot. They’re slightly ahead, but they have an army of 50+ phds and I don't see how a team like us can compete with them, that does not seem realistic... except that they're closed source. And we decided to…
What it does
In the maker’s words, at launch
Two months ago, my friends in AI and I asked: What if an AI could actually use a phone like a human? So we built an agentic framework that taps, swipes, types… and somehow it’s outperforming giant labs like Google DeepMind and Microsoft Research on the AndroidWorld benchmark. We were thrilled about our results until a massive lab (Zhipu AI) released its results last week to take the top spot. They’re slightly ahead, but they have an army of 50+ phds and I don't see how a team like us can compete with them, that does not seem realistic... except that they're closed source. And we decided to open-source everything. That way, even as a small team, we can make our work count. We’re currently building our own custom mobile RL gyms, training environments made to push this agent further and get closer to 100% on the benchmark. What do you think can make a small team like us compete against such giants? Repo’s here if you want to check it out or contribute: https://github.com/minitap-ai/mobile-use Our discord: https://discord.gg/6nSqmQ9pQs
Does the same job
all alternatives →
- DRDeep Research – Open-Source Customizable Reasoning Framework for Devs2025 · github.com · ▲7
The first ever AI peer reviewed research article just got approved. It’s kinda crazy how advanced AI have come to replace researchers. I've just been using Deep Research on ChatGPT and Perplexity a lot to write and research complex technical reports for my boss. He loves the reports and it has decreased my workload a ton but I still have some frustrations with it. None of them provide an API that gets me the same quality of output you would with the applications. I wanted something with more control on the LLMs, swappable with the reasoning new models that came out. Not just prompt →…
- BEBraintrust – Eval platform for AI products2023 · ▲8
Hey HN, We're excited to introduce Braintrust, a platform for running and tracking AI evaluations (“evals”) [1]. At my previous startup Impira and leading AI at Figma, we had this recurring problem where we never knew if changes we made to our products would improve or regress key user scenarios. We built some tooling to solve this problem and after talking to other developers learned that it was a widespread issue. Specifically, it’s challenging to establish a great dev loop that lets you systematically improve and ship high quality AI products. We worked with the teams at Zapier, Coda, and…
- ATA tool that turns everyday computers into your own AI cloud2024 · github.com · ▲5
I have a favour to ask. I’ve been working for a while on Kalavai, a project to make distributed AI easy. There are brilliant tools out there to help AI hobbyists and devs on the software layer (shout out to vLLM and llamacpp amongst many others!) but it’s a jungle out there when it comes to procuring and managing the necessary hardware resources and orchestrating them. This has always led me to compromise on the size of the models I end up using (quantized versions, smaller models) to save cost or to play within the limits of my rig. Today I am happy to share the first public version of our…
- IBI built cracked engineers – a new platform for technical job roles only2024 · crackedengineers.com · ▲8
I've built a job platform for technical roles called "Cracked Engineers". If you want to land a job with some of the world's best AI/tech startups - my goal is to make this the go-to place for that. YouTube walk-through can be found here: https://www.youtube.com/watch?v=XmuIOdES7mQ A huge thank you to Weaviate (Bob van Luijt & team), Unsloth AI (Daniel Han & Mike), OpenPipe (David Corbitt), Julius AI (Rahul Sonwalkar), Jimini AI (Luis F Voloch) for using the platform while in beta and posting first paid posts. So, why did I build it? Well, over the past period a lot of…
- YAYorph AI – a data engineer in your pocketNov 2025 · yorph.ai · ▲6
Hey folks! I'm a founding engineer at Yorph AI, an agentic data platform, built using ADK, that helps users (starting with product managers and analysts) join data from different sources (upload or sync), build version-controlled and reliable data workflows, and clean, analyze, and visualize data — all in one place. We're also releasing semantic layer creation later this week. The beta is live at yorph.ai/login — would love to hear your thoughts and feedback! (FYI: We're still waiting on Google app verification — you'll see a warning for a few days. Dropbox shows a similar one since…
More ai this month
the category →
I trained a 125M-parameter transformer to autocomplete piano performances in real time (~108 notes/sec on an iPhone 15). The idea is basically GitHub Copilot or Tabnine, except instead of prompting it with code, you prompt it by playing a few notes on a MIDI piano. The model then continues what you played, entirely on-device. The app is free if anyone wants to try it. Happy to answer questions about the model, training, Core ML, or the many things that didn't work.
AI · 17d ago · simedw.com
Astute▲585Automate your B2B brand going viral, with new media creators
AI · 18d ago · company-app.joinastute.com
Hey HN, Henry from Cactus here! We previously released Cactus Needle, a 14MB agentic LLM for tool call, device use, and structured extraction for phones, wearables, smart homes, small robots and microcontrollers. We got really great feedback here, and have now incorporated the suggestions to release Needle 2. The whole model is a single 14MB binary that runs a full session in 28MB of RAM; 45m parameters at 2bit compression. Needle hits 500 tokens/sec decode speed on a Raspberry Pi 5, sits between 400-1,500 tokens/sec on VR devices like Meta Quest 3S and Apple Vision Pro, and ranges…
AI · 27d ago · cactuscompute.com
Launched alongside, August 2025
the whole month →
- IS
I built the world's most impractical 1000-pixel display and anyone in the world can draw on it. It draws a single pixel at a time and takes 30-60 minutes to complete a single image. Anyone can participate in the project by voting for the next image to be drawn, and submitting images. https://kilopx.com/
Work · 2025 · benholmen.com
- KT
Kitten TTS is an open-source series of tiny and expressive text-to-speech models for on-device applications. We are excited to launch a preview of our smallest model, which is less than 25 MB. This model has 15M parameters. This release supports English text-to-speech applications in eight voices: four male and four female. The model is quantized to int8 + fp16, and it uses onnx for runtime. The model is designed to run literally anywhere eg. raspberry pi, low-end smartphones, wearables, browsers etc. No GPU required! We're releasing this to give early users a sense of the latency and voices…
Dev tools · 2025 · github.com
- IW
I was wondering how I can arrange objects along a spherical helix path, and read some articles on it. I ended up learning about parametric equations again, and make this visualization to document what I learned: https://visualrambling.space/moving-objects-in-3d/ feel free to visit and let me know what you think!
Life & fun · 2025 · visualrambling.space
- TC
For HTML Day 2025 [1], I made a web service that displays the current sky at your approximate location as a CSS gradient. Colours are simulated on-demand using atmospheric absorption and scattering coefficients. Updates every minute, without the use of client-side JavaScript. Source code and additional information is available on GitHub: https://github.com/dnlzro/horizon [1] https://html.energy/html-day/2025/index.html
Dev tools · 2025 · sky.dlazaro.ca