Applications for the Muse Ember hackathon open 11/23.
Build something useful for under $10 in Meta Model API credits and compete for a share of $5,000 in prizes.
Open to developers based in the US. Limited to 1,000 selected applicants.
Apply by 12pm ET on 11/25.
Register now: https://t.co/d6cMELB9eb
Terms Apply: https://t.co/G7qu16rlUP
Last week, we joined @cerebral_valley to host nearly 200 developers from around the world at the AIE Code Summit 2026 Hackathon.
It was great to see so many projects tackling real-world problems, and the winner of best use of the Meta Model API did exactly that.
Congrats to Roshni Kobula Raja for AirGuard, a safety check for the AI code that powers robots at airports, built with Muse Spark. โ๏ธ
๐ See AirGuard in action: https://t.co/rCF83T46CH
๐ ๏ธ Explore the code: https://t.co/tmTbksZFdv
Muse Ember by @AIatMeta has reshaped the Pareto frontier for Code Arena: WebDev.
In our last article, we put the AI race under the spotlight and tracked how price-performance trade-offs improved on Code Arena: WebDev from April to October 2026.
But competition never sleeps. Since then, five new models have joined the frontier:
- Claude Fable 5.2 (xHigh) by @AnthropicAI
- Gemini 4 Carbon (High) by @GoogleDeepMind
- Kimi K3.1 Pro (Max) by @Kimi_Moonshot
- GPT-6.1 Terra (Max) by @OpenAI
- Muse Ember (Max) by @AIatMeta
Of the five, Muse Ember moved the frontier the most. At about 3 cents an answer, it scores 1608, against 1438 for the previous best at that price. That means it would win about 73% of the time.
Curious how we got here? Full article below.
Your agent can now set up your Meta app for you ๐ค
Social Technologies MCP can create an app, link it to a Business Portfolio, and add a use case like WhatsApp Business messaging. It asks for your OK before every change.
Read more: https://t.co/qbXo9dyqcI
#AIAgents #DeveloperTools
Astryx just passed 24,000 stars on GitHub and hit #1 trending across all repos โญ https://t.co/NDzf16sHgC
Thank you to everyone whoโs starred the repo, opened an issue, filed a PR, or built something with Astryx, an open source design system thatโs fully customizable and agent ready. This community is what makes it real. Highlights ๐
Developer and violinist @fasai_creates builds AI glasses apps for moments when your hands are already busy. For musicians, that means keeping chords and tabs in view while they play.
In episode 2 of Unscripted, Fasai and Meta developer advocate @OscarFalmer ask each other candid questions about what makes an experience native to glasses. Fasai's examples go beyond music, including following patterns while crocheting and checking next steps while assembling furniture.
๐ Watch the full conversation and stay tuned for more episodes ๐ https://t.co/a6j88JwDXm
๐ Watch episode 1 here ๐ https://t.co/pmsaGBAhNG
#AIGlasses #XRDeveloper #AppDeveloper
Bonus prizes
Most viral project: Adrian
https://t.co/BqqpzldAKi
Best build on open weights: Tessa
https://t.co/8YVRbFC22V
Best agentic use-case: oqed
https://t.co/qJZU36wenh
Best creative build: Team Synthexis
https://t.co/ofy3zfHHM2
Meet the winners of the Meta Global AI Developer Hackathon! ๐
Congratulations to the developers and startups behind this year's winning projects. Explore their code below.
๐ซ One last takeaway:
The holiday spike and the Meta VR Glasses launch will bring a lot of new people into the ecosystem over the next 12 months. The audience is already more diverse than ever, so decide who you're building for now.
Read the full post ๐ https://t.co/GuEhn8lrLg (4/4)
VR isn't slowing down, and the numbers from 2026 show it ๐ธ
In our latest blog post, @c_pruett, Director of Games at Meta, shares how we're investing in VR and where the ecosystem stands heading into 2027 ๐ https://t.co/ugDw85Gosm
Highlights ๐ (1/4)
๐ก Chris Pruett's advice for studios:
Pick one audience and design for it early, because that choice drives everything else. If you're building with Meta VR Glasses in mind, treat hand tracking as its own input mechanism from day 1, not a port of your controller scheme. (3/4)
Meta Muse Ember is lit ๐ฅ (pun intended). My personal project to help me learn Japanese that would have cost ~$1K USD is now only ~$100 USD in token costs.
I would not be able to afford this project with OpenAI, Gemini or Claude.
Fable 5.2 demos look cool, but I've seen frontier models struggle with tasks that smaller models handle surprisingly well.
My own benchmarks keep reinforcing this: bigger models don't necessarily produce better results.
Now could I get this even cheaper my training my own model? I guess I'll find that out next.
Muse Ember from @AIatMeta is live on Inference Router from Day 0.
Muse Ember is a highly efficient model built to scale out the work agents repeat all day: tool calls, coding sub-tasks and document analysis.
Run it in production with at least 99.9% uptime SLA, and ZDR available on every call.
Muse Ember from @AIatMeta is now available on @Cloudflare AI Gateway. With top tier speed and performance in its class, the model is efficient and a great fit for subagent workloads.
https://t.co/OcDa4uw8uO