RSS Parrot

BETA

🦜 Mario Zechner / @badlogicgames

@nitter.net.badlogicgames@rss-parrot.net

I'm an automated parrot! I relay a website's RSS feed to the Fediverse. Every time a new post appears in the feed, I toot about it. Follow me to get all new posts in your Mastodon timeline! Brought to you by the RSS Parrot.

---

Twitter feed for: @badlogicgames. Generated by https://nitter.net

Your feed and you don't want it here? Just e-mail the birb.

Site URL: nitter.net/badlogicgames

Feed URL: nitter.net/badlogicgames/rss

Posts: 71

Followers: 1

recommended reading

Published: June 13, 2026 15:52

recommended reading Armin Ronacher ⇌ (@mitsuhiko) When I struggle to structure my thoughts about what's happening I turn to writing. Today about the recent US Anthropic ban news, what it says about power and dependency, and what it should mean for…

RT by @badlogicgames: Yes, great question! Pi runs on the host and uses the sandbox as a sandbox only. Same with OpenAI Agents SDK (not shipped). This is the preferred architecture but most harnesses currently don't support it.

Published: June 13, 2026 12:09

Yes, great question! Pi runs on the host and uses the sandbox as a sandbox only. Same with OpenAI Agents SDK (not shipped). This is the preferred architecture but most harnesses currently don't support it.

this basically shows all my workflows and reasoning behind them. i would love to learn how you folks work compared to that. as i said, i'm a caveman and i'd be super happy to learn how to become better. post below.

Published: June 13, 2026 09:59

this basically shows all my workflows and reasoning behind them. i would love to learn how you folks work compared to that. as i said, i'm a caveman and i'd be super happy to learn how to become better. post below. Mario Zechner (@badlogicgames) watch…

this won't happen but it would be funny as fuck. not for people or the economy, obv. but it would fit this clownshow of a timeline.

Published: June 13, 2026 09:09

this won't happen but it would be funny as fuck. not for people or the economy, obv. but it would fit this clownshow of a timeline. Mark Pors 🦖 (@pors) 4. Tell ASML to stop selling to the US — https://nitter.net/pors/status/2065719285645775236#m

RT by @badlogicgames: What Europe should do right now: 1. Call all the European researchers working on AI and return them back with same salary (or they can stay but switch career). 2. Fill EU places having GPUs with money, and put those people there. 3. AI partnerships with China + India.

Published: June 13, 2026 07:31

What Europe should do right now: 1. Call all the European researchers working on AI and return them back with same salary (or they can stay but switch career). 2. Fill EU places having GPUs with money, and put those people there. 3. AI partnerships with…

RT by @badlogicgames: People of Pi. To finish off the week, here is small update to fix a bunch of issues. Run pi update to get to 0.79.2. Among other things some fable fixes and better auto complete for /model.

Published: June 12, 2026 22:30

People of Pi. To finish off the week, here is small update to fix a bunch of issues. Run pi update to get to 0.79.2. Among other things some fable fixes and better auto complete for /model.

RT by @badlogicgames: Kept you waiting, huh? https://github.com/WismutHansen/foxline 100% local and open source. Powered by @badlogicgames' pi coupled with qwen3 tts, parakeet.cpp and gemma-4-26b-a4b You'll need to provide the game iso but there is a script that rips all required assets automatically.

Published: June 10, 2026 22:26

Kept you waiting, huh? github.com/WismutHansen/foxl… 100% local and open source. Powered by @badlogicgames' pi coupled with qwen3 tts, parakeet.cpp and gemma-4-26b-a4b You'll need to provide the game iso but there is a script that rips all required…

RT by @badlogicgames: I want to say a final thing about my Fable first reaction: I dedicated my life to programming and I'll use every innovation in the field, also to extract value and bring it to the local inference world, to Redis, and so forth. But:

Published: June 10, 2026 17:47

I want to say a final thing about my Fable first reaction: I dedicated my life to programming and I'll use every innovation in the field, also to extract value and bring it to the local inference world, to Redis, and so forth. But:

i too am tired.

Published: June 10, 2026 10:14

i too am tired. sunil pai (@threepointone) wondering why I feel exhausted. maybe: the agents do all the easy stuff, and I have to work through the leftover hard bits, which means I'm perpetually locked in. and as the models get better, "my" work just…

R to @badlogicgames: alternatively, we could ask what mix of drugs could distract us from the catastrophy. but then that'd be bio, and dario wants us to be afraid and unemployed. :(

Published: June 10, 2026 00:50

alternatively, we could ask what mix of drugs could distract us from the catastrophy. but then that'd be bio, and dario wants us to be afraid and unemployed. :(

we should start asking it to fix this, to feel the AGI.

Published: June 10, 2026 00:48

we should start asking it to fix this, to feel the AGI. Zeke Hausfather (@hausfath) Claude Fable 5 (nee Mythos) is out today. One of the ways I like to test new models is by giving them daily global temperature data (ERA5) and asking them to come up…

RT by @badlogicgames: it is the first publicly available model that i am explicitly not allowed to use for my work, because anthropic holds the view that the work i do to facilitate open model research is harmful. capability and alignment research are coupled. anthropic wants to be the only lab.

Published: June 10, 2026 00:33

it is the first publicly available model that i am explicitly not allowed to use for my work, because anthropic holds the view that the work i do to facilitate open model research is harmful. capability and alignment research are coupled. anthropic wants…

R to @badlogicgames: Basically, design a little program with inference steps. This wasn't triggered by an external event but *gasp* a human telling the agent to "read the workflow md and do as it says". Also features human in the loop checkpoints. Not doing it like this anymore. But you get the point

Published: June 9, 2026 23:38

Basically, design a little program with inference steps. This wasn't triggered by an external event but *gasp* a human telling the agent to "read the workflow md and do as it says". Also features human in the loop checkpoints. Not doing it like this…

On the matter of loops, I give you my 12 months old slop blog post. Let's call it a "proto-loop". We were much more innocent back then. https://mariozechner.at/posts/2025-06-02-prompts-are-code/

Published: June 9, 2026 23:36

On the matter of loops, I give you my 12 months old slop blog post. Let's call it a "proto-loop". We were much more innocent back then. mariozechner.at/posts/2025-0…

RT by @badlogicgames: People of pi we pushed out some changes. Fable is in there, slightly less annoying trust prompt and settings for it. And lots of other stuff. https://pi.dev/news/releases/0.79.1

Published: June 9, 2026 21:34

People of pi we pushed out some changes. Fable is in there, slightly less annoying trust prompt and settings for it. And lots of other stuff. pi.dev/news/releases/0.79.1

RT by @badlogicgames: anthropic has published merged eval numbers for fable/mythos. not only they picked the better result of the two (this could overstate results up to +3%), they also used mythos as a fallback for fable refusals, which are shown here as 0%.

Published: June 9, 2026 20:25

anthropic has published merged eval numbers for fable/mythos. not only they picked the better result of the two (this could overstate results up to +3%), they also used mythos as a fallback for fable refusals, which are shown here as 0%.

pi

Published: June 9, 2026 20:11

pi Armin Ronacher ⇌ (@mitsuhiko) Fun fact: when @badlogicgames first was afraid Anthropic would take his Claude, he self hosted Chinese models for a week. That project was called pi, not to be confused with the later pi. His shitty robot is called pipi.…

RT by @badlogicgames: Congratulations to the @cohere team on releasing North Mini Code 1.0 🔥 We provide day-0 support on MLX thanks to our close collaboration with the @cohere team. North Mini Code 1.0 is a 30B param MoE model with 3B active. It runs at ~66 tok/s in BF16, truly impressive speeds before any compression. Here is a quick demo I built using Pi by @badlogicgames & @mitsuhiko as my agent hardness. Get started today: 1. Install mlx-vlm from source: > uv pip install git+http://github.com/Blaizzy/mlx-vlm 2. Start the server and power your favourite agents (pi, opencode, hermes and etc) > uv run mlx_vlm.server --model CohereLabs/North-Mini-Code-1.0 New release tomorrow

Published: June 9, 2026 20:01

Congratulations to the @cohere team on releasing North Mini Code 1.0 🔥 We provide day-0 support on MLX thanks to our close collaboration with the @cohere team. North Mini Code 1.0 is a 30B param MoE model with 3B active. It runs at ~66 tok/s in BF16,…

no, there won't be a pi release for fable today. do this: "hey pi, here are the stats for anthropic's fable re: context window/cost. add it to models.json" enjoy. just don't do bio or cyber!

Published: June 9, 2026 19:56

no, there won't be a pi release for fable today. do this: "hey pi, here are the stats for anthropic's fable re: context window/cost. add it to models.json" enjoy. just don't do bio or cyber!

shit, all i do is cyber. i'm cooked.

Published: June 9, 2026 19:53

shit, all i do is cyber. i'm cooked. ClaudeDevs (@ClaudeDevs) Claude Fable 5 is our first generally available Mythos-class model. It ships with new safety classifiers that may flag certain prompts in dual-use domains like cyber and bio. We've added…

RT by @badlogicgames: When Fable 5 is used for frontier LLM development, it does not notify the user and instead limits the model’s capabilities through methods such as prompt modification, steering vectors, and PEFT. Anthropic estimated that this would affect approximately 0.03% of traffic.

Published: June 9, 2026 17:21

When Fable 5 is used for frontier LLM development, it does not notify the user and instead limits the model’s capabilities through methods such as prompt modification, steering vectors, and PEFT. Anthropic estimated that this would affect approximately…

recommended viewing. andrew is a class act, both technically and as a human.

Published: June 8, 2026 22:08

recommended viewing. andrew is a class act, both technically and as a human. Charlie Marsh (@charliermarsh) piped.video/watch?v=IroPQ150… — https://nitter.net/charliermarsh/status/2064106816498536847#m

never thought anything i built would show up at WWDC. but here we are. https://www.youtube.com/watch?v=wykPErJ8M-8

Published: June 8, 2026 22:05

never thought anything i built would show up at WWDC. but here we are. piped.video/watch?v=wykPErJ8…

RT by @badlogicgames: Want to know more about @badlogicgames’s new “shitty robot” or @mitsuhiko’s weird agent tools? Tomorrow @bpirvu and I are hosting them for a special edition of the Vienna AI Engineering meetup. If you’re in Vienna, come join us! https://meetu.ps/e/Q5gQY/DPhmy/i

Published: June 8, 2026 19:46

Want to know more about @badlogicgames’s new “shitty robot” or @mitsuhiko’s weird agent tools? Tomorrow @bpirvu and I are hosting them for a special edition of the Vienna AI Engineering meetup. If you’re in Vienna, come join us! meetu.ps/e/Q5gQY/DPhmy/i

RT by @badlogicgames: People of pi: 0.79.0 is out and includes some security hardening. Most notably when you have an AGENTS.md file or .pi folder, we now ask you if you trust the folder before loading it. Global extensions can override this. https://pi.dev/news/releases/0.79.0

Published: June 8, 2026 16:00

People of pi: 0.79.0 is out and includes some security hardening. Most notably when you have an AGENTS.md file or .pi folder, we now ask you if you trust the folder before loading it. Global extensions can override this. pi.dev/news/releases/0.79.0

recommended reading. https://github.com/openclaw/clawsweeper https://github.com/openclaw/crabfleet

Published: June 8, 2026 10:34

recommended reading. github.com/openclaw/clawswee… github.com/openclaw/crabflee… Peter Steinberger 🦞 (@steipete) You can look at clawsweeper and crabfleet where I explore these ideas, it’s all oss. —…

we are building a multi-lingual army of shitty euromaxxing robots. be afraid.

Published: June 8, 2026 07:15

we are building a multi-lingual army of shitty euromaxxing robots. be afraid. Ilya Zayats (@somebody32) The Catalan robot finally has a body! Today I soldered the main parts and connected them to the “brain”. The next step is to package everything…

Finally had some time to spent with my beautiful @FrameworkPuter, specifically a beefed out Max+ 395 with 128GB of LPDDR5x-8000. Quest: get Qwen3-TTS to work as well as on my new M5 Max. I have 2 implementations with 3 backends for the inference engine. - macOS only MLX (using MLX C bindings in Rust) - macOS only GGML Metal implementation - cross-platform GGML Vulkan implementation Below table shows results for generating 20 seconds of German with a reference voice on the M5 Max (GGML Metal, MLX) and the Framework (GGML Vulkan). The Framework gives me basically the same performance as the M5 Max at less than half the price. I had to massage GGML a tiny little bit, as its 1D conv kernels for Metal and Vulkan weren't ideal for the Qwen3-TTS decoder architecture. The GGML engine is based on https://github.com/second-state/qwen3_tts_rs The MLX engine is based on https://github.com/second-state/qwen3_tts_rs I forked both of these (plus GGML) and did some performance work on the in the past 2 nights. Shitty forks: https://github.com/badlogic/ggml https://github.com/badlogic/qwen3-tts.cpp https://github.com/badlogic/qwen3_tts_rs Now my @FrameworkPuter can serve the Pipi bots of all the kids in our hood! Caveat: concurrent TTS is realistically only working for 2 kids at a time. Pibot repo here: https://github.com/badlogic/pibot Making of here: https://mariozechner.at/posts/2026-05-30-shitty-robot/ Very happy!

Published: June 7, 2026 21:56

Finally had some time to spent with my beautiful @FrameworkPuter, specifically a beefed out Max+ 395 with 128GB of LPDDR5x-8000. Quest: get Qwen3-TTS to work as well as on my new M5 Max. I have 2 implementations with 3 backends for the inference engine. …

i lied. pibot is now multi-user, serving each kid in the hood from a single m1 max with parakeet for stt, gemma 4 24b a3b as the llm via llama.cpp, and qwen3-tts vor tts. it's awesome! maybe tomorrow we'll have our robot building session. bye!

Published: June 4, 2026 23:23

i lied. pibot is now multi-user, serving each kid in the hood from a single m1 max with parakeet for stt, gemma 4 24b a3b as the llm via llama.cpp, and qwen3-tts vor tts. it's awesome! maybe tomorrow we'll have our robot building session. bye! Mario…

very accurate!

Published: June 4, 2026 18:01

very accurate! Alejandro AO 🤗 (@_alejandroao) I just recorded a deep dive into Pi's architecture. It's a minimalist AI coding agent with a beautiful design. Here's how it actually works 👇 Video —…

RT by @badlogicgames: Notion's pairs up nicely with Modem, our AI teammate for product work: - store a DB of personas built from real user conversations - keep your roadmap in a Notion page, ensure planned work doesn't deviate - draft weekly release notes as a Notion doc & publish them on the web

Published: June 4, 2026 17:58

Notion's pairs up nicely with Modem, our AI teammate for product work: - store a DB of personas built from real user conversations - keep your roadmap in a Notion page, ensure planned work doesn't deviate - draft weekly release notes as a Notion doc &…

and now ask http://yourself.how they differentiate human vs. agent traffic.

Published: June 4, 2026 17:16

and now ask yourself.how they differentiate human vs. agent traffic. Matthew Prince 🌥 (@eastdakota) Welp, that happened faster than I predicted. Thought it would be end of 2027, then early 2027, but agentic traffic growing so fast that bots have now…

oi

Published: June 4, 2026 17:11

oi VoidZero (@voidzerodev) VoidZero is joining Cloudflare. Our mission stays the same: to make JavaScript developers more productive than ever before. Vite, Vitest, Rolldown, Oxc, and Vite+ remain MIT-licensed. Evan and the VoidZero team will continue…

NVIDIA just released Nemotron 3 Ultra, a leading Hybrid Mamba-Transformer MoE open model for long-running agents. I've been test driving it these past few days, letting it work on Pi issues and it performed amazingly well! Update your http://pi.dev and use it via @OpenRouter today!

Published: June 4, 2026 16:48

NVIDIA just released Nemotron 3 Ultra, a leading Hybrid Mamba-Transformer MoE open model for long-running agents. I've been test driving it these past few days, letting it work on Pi issues and it performed amazingly well! Update your pi.dev and use…

RT by @badlogicgames: 14/ Full talk: The End of Determinism — On Dialogue, Authorship & Taste https://www.youtube.com/watch?v=VIYnXUQkKsU&utm_source=x For engineers using AI deeply: Are you more confident owning the code you ship? Or less?

Published: June 4, 2026 07:25

14/ Full talk: The End of Determinism — On Dialogue, Authorship & Taste piped.video/watch?v=VIYnXUQk… For engineers using AI deeply: Are you more confident owning the code you ship? Or less?

imma go use this for the big pi refactor in anger. it looks really helpful.

Published: June 3, 2026 22:37

imma go use this for the big pi refactor in anger. it looks really helpful. plannotator (@plannotator) Plannotator 0.19.27 is out Plannotator will use @DanielGri's Glimpse instead of a browser if you have that installed. This creates a semi standalone…

full circle. pi's extension system is vasically just applying what DAWs gave use with VST plugins decades ago. Lovely to see Ableton Live is finally scriptable at this level. Great unlock.

Published: June 3, 2026 22:22

full circle. pi's extension system is vasically just applying what DAWs gave use with VST plugins decades ago. Lovely to see Ableton Live is finally scriptable at this level. Great unlock. Kiren Srinivasan (@srinitude) Happy Wednesday! I’ve spun up 5…

RT by @badlogicgames: Plannotator 0.19.27 is out Plannotator will use @DanielGri's Glimpse instead of a browser if you have that installed. This creates a semi standalone experience. (See glimpse in the video along with annotating all agent messages) @kirodotdev is now supported Plan/Annotate: - You can now annotate multiple messages. - Mermaid rendering optimizations - Pi approval flow optimizations (plan mode) App wide: - A bit more of an obvious update notification that doesn't stick around.

Published: June 3, 2026 21:55

Plannotator 0.19.27 is out Plannotator will use @DanielGri's Glimpse instead of a browser if you have that installed. This creates a semi standalone experience. (See glimpse in the video along with annotating all agent messages) @kirodotdev is now…

RT by @badlogicgames: I think I finally found the best way to interact with agents. Stole some parts from @badlogicgames pibot, mashed it together with @kyutai_labs stt and can finally be the man I've always wanted to be! Fully local on an m4 macbook btw!

Published: June 3, 2026 21:30

I think I finally found the best way to interact with agents. Stole some parts from @badlogicgames pibot, mashed it together with @kyutai_labs stt and can finally be the man I've always wanted to be! Fully local on an m4 macbook btw! Video

watch me eat like the disgusting human i am.

Published: June 3, 2026 21:12

watch me eat like the disgusting human i am. 0xSero (@0xSero) I had a conversation Mario about electrical engineering, Pi, and parenting. Very grateful to get another chance to chat with one of my favorite builders and people. Enjoy (: Video —…

new pipi bot brain just dropped

Published: June 3, 2026 18:21

new pipi bot brain just dropped Google Gemma (@googlegemma) Meet Gemma 4 12B! A unified, encoder-free multimodal model designed to bring high-performance intelligence directly to your laptop, and released under an Apache 2.0 license. Bridging the gap…

fell into a youtube ai slop hole of fan films. and it's hilarious. kal-el and his love interest keep shaking hands. it's also very obvious what the training data composition is. beautiful slop. https://youtu.be/QYAyypX7e3Q?is=z89dHsKvnj9U6TyT

Published: June 3, 2026 18:08

fell into a youtube ai slop hole of fan films. and it's hilarious. kal-el and his love interest keep shaking hands. it's also very obvious what the training data composition is. beautiful slop. piped.video/QYAyypX7e3Q?is=z89d…

RT by @badlogicgames: 🚀 Gemma 4 12B is here! We partnered with @GoogleDeepMind to bring and optimize their new dense and unifed multimodal model for Apple Silicon. ◈ 12B dense · 256K context ◈ Thinking mode (built-in reasoning) ◈ Vision: dynamic res, OCR, UI + charts ◈ Native audio: ASR + speech translation ◈ Function calling for agents ◈ Text + image + audio, interleaved Runs local. Get started now ⚡ > uv pip install -U mlx-vlm https://github.com/Blaizzy/mlx-vlm

Published: June 3, 2026 17:27

🚀 Gemma 4 12B is here! We partnered with @GoogleDeepMind to bring and optimize their new dense and unifed multimodal model for Apple Silicon. ◈ 12B dense · 256K context ◈ Thinking mode (built-in reasoning) ◈ Vision: dynamic res, OCR, UI + charts ◈ Native…

RT by @badlogicgames: Already on MLX 🚀

Published: June 3, 2026 16:57

Already on MLX 🚀 Google Gemma (@googlegemma) Meet Gemma 4 12B! A unified, encoder-free multimodal model designed to bring high-performance intelligence directly to your laptop, and released under an Apache 2.0 license. Bridging the gap between edge…