Conductor AI Labs

Small team dedicated to making cool stuff.

[Conductor AI Labs @conductorailabs]

QR codes look boring.

So we made them look fun!

Check out these animated, video, and 3D QR codes 🎨

Change you animation style or colours, upload you own video, then save out working QR codes that are wildly vibrant and super funky.

You can save out pngs, webm, and gif formats, plus select your size, clip length, and adjust for speed or timing.

Put them across your web and social properties, or in your videos for surprising and delightful consumer engagements.

Try it for yourself ⬇️

Site: https://t.co/O7sl8162fU
GitHub: https://t.co/kdwsg6yvMK

[Brian Johnson @_brian_johnson]

@conductorailabs @waveshare00 The hidden usage split is the interesting part. Once cache reads and thinking show up separately, the tracker becomes way more useful than a raw output counter.

[Conductor AI Labs @conductorailabs]

Our latest experiment. Terminal token tracker on a USB screen.

What worked, what was a challenge, and what was learned about hidden usage:

What worked:
-Getting the @waveshare00 USB screen display working.
-Green +tokens are what the agent displays in chat.
-Blue +tokens are what are hidden but still go toward your limit.
-Rainbow lights every time your tokens increase 🤣
-Real time tracking on every message (through Claude subscription).
-Trackers covering 5h session limit and Weekly limit.
-Display total tokens spent that day.

What was a challenge:
-Understanding how tokens are spent per message and hidden token costs.
-Ensuring weekly totals remained matched to /usage.
-Deciding what cool data to leave off for space.
-Showing it work in real-time next to a screen for this video.

What was learned about hidden token use:
-Claude says x tokens were output, but more 'usage' is happening.
-Cache read and write take up a lot of token usage per message.
-Thinking is not included in the output total it seems.
-There are a lot of token-saving tactics and it's an evolving thing

Overall, it was a lot of fun working on this little prototype and a lot was learned about how tokens are used through Claude subscription.

Going to continue experimenting with this cool little screen, it actually has an S3 on it, which is a pretty sweet starting ground for creative fun.

Tally-ho!

Want to talk about making cool stuff? Reach out and follow @conductorailabs along the way for more hypothetical fun.

#experiment #ai #claude #creative #tokens #prototype #hardware #waveshare #esp32 #s3 #app #smart #development #experiment #budget #management #business #tech #education #motivation

[Conductor AI Labs @conductorailabs]

Qwen3.6-27B is out, the latest dense, open-source model, packing flagship-level coding power!

27B.

What's new:
🧠 Outstanding agentic coding — surpasses Qwen3.5-397B-A17B across all major coding benchmarks
💡 Strong reasoning across text & multimodal tasks
🔄 Supports thinking & non-thinking modes
✅ Apache 2.0

[Qwen @Alibaba_Qwen]

🚀 Meet Qwen3.6-27B, our latest dense, open-source model, packing flagship-level coding power!

Yes, 27B, and Qwen3.6-27B punches way above its weight. 👇

What's new:
🧠 Outstanding agentic coding — surpasses Qwen3.5-397B-A17B across all major coding benchmarks
💡 Strong reasoning across text & multimodal tasks
🔄 Supports thinking & non-thinking modes
✅ Apache 2.0 — fully open, fully yours

Smaller model. Bigger results. Community's favorite. ❤️
We can't wait to see what you build with Qwen3.6-27B! 👀

🔗👇
Blog: https://t.co/P2Zx7FwMxB
Qwen Studio: https://t.co/c4vm4LuZrU
Github: https://t.co/zKDEbv0R4U
Hugging Face:
https://t.co/N67hyzxvfr
https://t.co/SSdtbWRDap
ModelScope:
https://t.co/xODf1pj9kw
https://t.co/xXhoqlJ2AB

[Conductor AI Labs @conductorailabs]

Need some music for marketing your new product?

Stan Bran is your man.

Simply tell him what's so groovy about your business or service, sit back, and watch the audio magic happen.

https://t.co/hxlKxR0MAa

#elevenhacks #turbopuffer #funny @ElevenLabs
@turbopuffer

[Conductor AI Labs @conductorailabs]

Knowladex outputs a really clean, easy to digest wiki format that both players and developers will love.

#gamedev #indiedev #indiegame #AI #MCP #Obsidian

[Conductor AI Labs @conductorailabs]

One thing Knowladex is really useful for is creating and updating a complete video game wiki with just a few clicks. It's incredibly easy and effective.

🎮 Game Files > 🤖 MCP Server > 🗂️ Knowladex Public URL

🔗 Public Wiki Example: https://t.co/uMNsbSxv6b

#gamedev #mcp #knowledgebase #corpus #AI #gaming #videogame #wiki #development

[Conductor AI Labs @conductorailabs]

Introducing Knowladex.

Knowledge bases that write themselves.

Powered by an MCP server via agent, or simple drag and drop, teams can organize their documents into an incredibly helpful wiki-style knowledge base.

In the video, you'll see the result of asking the agent to ingest the Mythos system card, and compiled into helpful cross-linked documents complete with charts and images.

⭐️ Share your wiki with an optional public URL
⭐️ Export to Obsidian or other platforms
⭐️ Use the same MCP across orgs

Beta opening soon. Save your spot.

🔗https://t.co/b6xuJysIsK

@karpathy #AI #Wiki #Obsidian #MD #Helpful

[Conductor AI Labs @conductorailabs]

@karpathy @chalish_b @kepano Our Knowladex uses an MCP server to ingest the Mythos PDF, so anyone on our team can do it remotely to the same org, and it turned out great. We even make the URL's optionally shareable so you can outsource your knowledge base.
https://t.co/zyJJUKSRLq

[Conductor AI Labs @conductorailabs]

Introducing Knowladex.

Knowledge bases that write themselves.

Powered by an MCP server via agent, or simple drag and drop, teams can organize their documents into an incredibly helpful wiki-style knowledge base.

In the video, you'll see the result of asking the agent to ingest the Mythos system card, and compiled into helpful cross-linked documents complete with charts and images.

⭐️ Share your wiki with an optional public URL
⭐️ Export to Obsidian or other platforms
⭐️ Use the same MCP across orgs

Beta opening soon. Save your spot.

🔗https://t.co/b6xuJysIsK

@karpathy #AI #Wiki #Obsidian #MD #Helpful

[Conductor AI Labs @conductorailabs]

Introducing Knowladex.

Knowledge bases that write themselves.

Powered by an MCP server via agent, or simple drag and drop, teams can organize their documents into an incredibly helpful wiki-style knowledge base.

In the video, you'll see the result of asking the agent to ingest the Mythos system card, and compiled into helpful cross-linked documents complete with charts and images.

⭐️ Share your wiki with an optional public URL
⭐️ Export to Obsidian or other platforms
⭐️ Use the same MCP across orgs

Beta opening soon. Save your spot.

🔗https://t.co/b6xuJysIsK

@karpathy #AI #Wiki #Obsidian #MD #Helpful

[Conductor AI Labs @conductorailabs]

Creating visualized stats our citizens (each a unique model) generate every world.

Every citizen now has it's own proto-voice thanks to @ElevenLabs Voices, all running on @Replit

Working on more visualizations for data and consolidating between worlds.

#ElevenHacks #AI #Replit #ElevenLabs #Creative #LLM

[Conductor AI Labs @conductorailabs]

If the world as we know it ends, what does starting over from scratch look like?

We're running the experiment. A work in progress.

30 individual AI language models. One world. No starting knowledge, just pure instinct. Each model retrained to make better decisions from the previous run.

Hear their proto-language thanks to @ElevenLabs and see their progress thanks to @Replit

https://t.co/yNIJlYyTqv

#AI #ElevenHacks #Creative #Hackathon #Research

[Conductor AI Labs @conductorailabs]

One thing we've been working on and struggling with is the balance between hardware demands and user experience.

Every citizen has their own voice and thought process that evolves.

So, when their current world ends, we additively train all 30 models with enhanced decision making abilities to prepare them for the new world.

It takes time and a lot of monitoring/health checking.

Still a fun experiment and pushing the limits.

#AI #FineTune #LocalModel #Train #ElevenHacks @ElevenLabs @Replit

[Alex Albert @alexalbert__]

We released Claude Opus 4.6 just two months ago. Today we're sharing some info on our new model, Claude Mythos Preview.

[Anthropic @AnthropicAI]

Introducing Project Glasswing: an urgent initiative to help secure the world’s most critical software.

It’s powered by our newest frontier model, Claude Mythos Preview, which can find software vulnerabilities better than all but the most skilled humans.
https://t.co/NQ7IfEtYk7

[Conductor AI Labs @conductorailabs]

If the world as we know it ends, what does starting over from scratch look like?

We're running the experiment. A work in progress.

30 individual AI language models. One world. No starting knowledge, just pure instinct. Each model retrained to make better decisions from the previous run.

Hear their proto-language thanks to @ElevenLabs and see their progress thanks to @Replit

https://t.co/yNIJlYyTqv

#AI #ElevenHacks #Creative #Hackathon #Research

[Conductor AI Labs @conductorailabs]

We gave your houseplants a voice.

Because millions of people claim they can't keep a plant alive, no matter what they try.

Now they can try talking to them.

Get real-time voice reports from your houseplants just by asking, tapping, or texting.

https://t.co/cPPsek2tQU

#ElevenHacks @ElevenLabs @Cloudflare @hgtv #gardening @Gizmodo @TechCrunch @BHG @_houseandgarden @RobotShop #funny #gardening #spring #planting #flowers

[Conductor AI Labs @conductorailabs]

@karpathy Have a similar workflow since experimenting with novels containing continued and intertwining storylines.

We set up an MCP server so we can add and pull from any knowledge base as we go. Added in a few helpful tools and it's really smooth, plus works remotely.

[Conductor AI Labs @conductorailabs]

We gave your houseplants a voice.

Because millions of people claim they can't keep a plant alive, no matter what they try.

Now they can try talking to them.

Get real-time voice reports from your houseplants just by asking, tapping, or texting.

https://t.co/cPPsek2tQU

#ElevenHacks @ElevenLabs @Cloudflare @hgtv #gardening @Gizmodo @TechCrunch @BHG @_houseandgarden @RobotShop #funny #gardening #spring #planting #flowers