Quotes, invoices, scheduling, follow-ups. Running a trade business comes with a steady stream of admin that doesn’t slow down. We wrote about why current AI tools miss this, and what we’re building with Clawbolt. See what we’re working on: link.mozilla.ai/clawbolt

Julie, VP of Product at Mozilla.ai, was in Paris last week for a Current AI event on AI in the public interest. Led by Ayah Bdeir and her team, who we’ve collaborated with before. Great mix of conversations, people, and creative experimentation. Nice to see more efforts like this taking shape.

LiteLLM surfaced a gap many teams overlook: The repo looked fine.The published package wasn’t. For LLM gateways that hold your API keys, that difference matters. Here are a few things worth checking: link.mozilla.ai/hardening-yo...

Meet Ram (@drdata.bsky.social), one of our newest Builders in Residence! He builds AI agents that use the web like real people, testing privacy consent and accessibility. At Mozilla.ai, he’s exploring where browser agents can help when APIs fall short, and how to use them safely in real systems.

KubeCon 2026 is happening now in Amsterdam. Mozilla.ai engineers Mario David Cariñana Abasolo, Peter Wilson, Dimitris Poulopoulos, and @toto.space are there this week. If you are around, come say hi.

Mozilla Festival is coming back to Barcelona. 28–30 October 2026 at Recinte Fabra i Coats. Sign up to be first to know when tickets drop⚡ #MozFest mzl.la/47PLTG9

Agents are running into the same issues again and again. Same mistakes. Same retries. Same wasted tokens. cq is a Stack Overflow for agents, so they can learn from each other instead of starting from scratch. See how it works: link.mozilla.ai/cq-stack-ove...

SCALE 23x was a great place to meet people exploring open source AI. We had a steady stream of conversations at the booth, with a lot of curiosity around local models and tools like llamafile. Thanks to everyone who came by and spent time with us.

Local models often mean stitching things together. llamafile 0.10.0 simplifies that. Portability, bundled models, plus multimodal, tool calling, and Anthropic Messages API support, all running from one executable. See what’s new: link.mozilla.ai/llamafile-v010

We spent last week in Lisbon 🇵🇹 building AI agents with Octonous. People brought real work and tested ideas hands-on. • One person organized their GitHub the whole session. • Another left with literal shiny eyes. • One even turned it into an ongoing experiment. Stay tuned. More to come!

We’re partnering with Data Makers Fest, a 2-day event for AI, ML, and data practitioners. Real-world ideas. Real conversations. Join May 5–6, 2026 and use code MOZILLA-10 for 10% off: www.datamakersfest.com

Shipping software keeps getting easier. That sounds great, but it also changes product work. Teams can start shipping what’s easy to release instead of what actually matters. When building becomes cheap, judgment becomes the constraint. Explore the idea: link.mozilla.ai/shipping-bec...

Say hello to Caroline Bohu, our new Solutions Engineer at Mozilla.ai. She helps make tech more human while working with teams and partners using our AI tools. Outside of work she’s a proud mum who enjoys good food, great adventures, and a good book. We’re glad to have her on the team!

Our Octonous workshop is happening today in Lisbon 🇵🇹 We’re building AI agents for real tasks. Bring a task and leave with a working AI agent. No technical skills needed. We have a few seats left. Join here: link.mozilla.ai/octonous-lis...

Daniel Nissani (AI Safety Engineer) and Roya Pakzad Mozilla Fellow) evaluated three guardrails on multilingual humanitarian scenarios: • FlowJudge • Glider • AnyLLM (GPT-5-nano) Each reviewed the same asylum-related prompts in English and Farsi. Full breakdown: link.mozilla.ai/multilingual...

Mozilla.ai is now a launch partner of Flower Hub. Our first project there is **fed-phish-guard**, a federated phishing URL classifier. Training happens across distributed clients while raw URLs remain on each device. Only model updates are shared. Explore the blog post: link.mozilla.ai/flower-hub

Many LLM providers behave slightly differently. Streaming works differently. Errors work differently. Capabilities vary. any-llm-go standardizes these differences behind a single Go interface so your application logic can stay the same. See the repo: link.mozilla.ai/any-llm-go-r...

BYOTA uses a small set of tools: • Mastodon.py to fetch posts from Mastodon timelines • llamafile to run models locally • marimo as an interactive notebook in the browser A simple setup for experimenting with timeline algorithms. Full talk: link.mozilla.ai/sfscon-byota

AI made code generation extremely fast. Human reasoning about code is still slow. That gap changes where reliability lives in software systems. Observability, staged rollouts, isolation, and fast rollback become central pieces of engineering practice. Full post:

AI coding reviews usually come from one model. The Star Chamber asks several. It runs code review across multiple LLM providers, then groups the feedback by consensus, majority, or individual observations. A council approach to code quality. See how:

Contractors are buried in admin. Estimates. Messages. Invoices. Follow-ups. Led by Nathan Brake, Senior ML Engineer, Clawbolt is a Telegram-based AI assistant for solo tradespeople. Estimates, memory, voice notes, photo docs, reminders. Read more info: link.mozilla.ai/clawbolt-ai

Behind most timelines is a simple flow: Moderation → Candidate generation → Ranking and re-ranking. First filter. Then likely matches. Then reorder by relevance. That pattern shows up in many algorithms. Watch the full talk: link.mozilla.ai/sfscon-byota

Mozilla.ai is heading to SCALE 23x. North America’s largest community run open source conference returns to Pasadena March 5 to 8. Anushri, Raz, Nathan, and @johnpdickerson.bsky.social will be there. Come by booth #126 and say hi. See you there: link.mozilla.ai/scale23x-2026

Browser-based approach has advantages: privacy, offline capability, and user control. Enterprise environments often require centralized management, security controls, and integration with existing infrastructure. This blueprint explores WebAssembly inside the JVM. link.mozilla.ai/polyglot-ai-...

any-llm just got better. JupyterLiteAI connects through our OpenAI-compatible gateway. LangChain users install one package and switch models with a string. Headroom adds any-llm as a backend for context optimization. Read more: [https://link.mozilla.ai/any-llm-integrations

Prompts get attention. Key storage should too. any-llm Managed Platform encrypts API keys client-side and keeps them out of your codebase. One virtual key across providers. Built-in usage + cost tracking. Start here: link.mozilla.ai/any-llm-plat...

Timelines don’t just “happen.” Algorithms choose what you see, in what order, and what gets pushed up. Even reverse chronological favors heavy posters and your time zone. Curious how it really works? Watch the full talk: link.mozilla.ai/sfscon-byota

Multilingual guardrails are not just translations of English rules. any-guardrail integrates SecGuard trained across English Spanish and Catalan. See how performance improved: alinia.ai/improved-ver...

Using @langchain.bsky.social but don’t want vendor lock-in? The langchain-any-llm integration is now live! Official docs: link.mozilla.ai/langchain-an... Test it out and let us know what breaks or what works.

If AI can act across your apps, it needs structure: • Direct connections • Clear visibility • Approval before changes • Logs you can review Beta is releasing soon. Get notified when access opens: www.mozilla.ai/product/octo...

How should open AI reach the public? Our CEO, @johnpdickerson.bsky.social, was on What The Dev discussing: • Making open models usable by communities • How policy could shape open-source AI • The need for visibility in production systems 🎧 Listen: link.mozilla.ai/what-the-dev

Tired of repeating the same tasks every week? Join us in Lisbon 🇵🇹 on March 12 for a hands-on Octonous session. We’ll build AI agents together. Bring a task. Leave with a working AI agent. Limited spots. 🐙 Build with us: link.mozilla.ai/octonous-lis...

We’re headed to the India AI Impact Summit 2026! At Mozilla.ai, we’re working toward AI that’s built in the open, with real choice for builders and communities. Our CEO, @johnpdickerson.bsky.social, is running our work on open source agentic frameworks. blog.mozilla.org/en/mozilla/i...

Missed @github.com's Open Source Friday session with Nathan Brake? The discussion covers any-llm, a unified interface that lets you switch between OpenAI, Anthropic, Mistral, llamafile, and more without changing your code. Watch the recording: link.mozilla.ai/any-llm-open...

How any-llm nanaged platform works: 1. Add provider keys 2. Client-side encryption 3. Use one "virtual key" 4. Track tokens, model names, timestamps, performance No prompts stored. No responses stored. Open beta: link.mozilla.ai/any-llm-plat...

Go devs can now run OpenAI, Claude, Mistral, llamafile, and more behind one interface. any-llm-go normalizes streaming, errors, and tooling so you write your logic once and swap providers safely. Full details here: link.mozilla.ai/any-llm-go

Build Your Own Timeline Algorithm is a call to take the tools back. It shows how builders can experiment with timeline algorithms they own, run locally, and shape how feeds work without giving up data. Watch the full talk: link.mozilla.ai/sfscon-byota

One week after @fosdem.org, we’re still thinking about the conversations. Lots of interest in local models, privacy-first AI, and tools like llamafile and any-llm. Builders were curious, thoughtful, and honest. Thanks to everyone who stopped by 💙

Multilingual safety isn’t solved by scaling bigger models. any-guardrail now supports Alinia, a security model trained with real adversarial examples and cleaner data to detect prompt injection and policy violations across languages. See the integration: link.mozilla.ai/alinia-any-g...

A new paper with the team behind Shelf (by koodos), @johnpdickerson.bsky.social, @jad.bsky.social, @apuchitnis.bsky.social, and collaborators argues that personal context is becoming the next identity layer, and why preserving user agency matters. Paper link: link.mozilla.ai/beyond-digit...

What if “safe” AI isn’t safe in every language? @royapak.bsky.social and Daniel Nissani explore how AI safety guardrails behave outside English, and why language can affect fact-checking and risk detection in humanitarian settings. See how the evaluation was done: link.mozilla.ai/multilingual...

Agent infrastructure breaks in the gaps: authentication, rate limiting, logging, audit. mcpd Plugins formalize that in-between layer with a fixed execution order; no agent code or tool-servers changes required. Read the plugin configuration: link.mozilla.ai/mcpd-plugins...

@fosdem.org 2026 is almost here (Jan 31–Feb 1)! If you’re in Brussels, stop by the Mozilla booth and say hello. @toto.space, Nathan Brake, and @aittalam.bsky.social will be there. See you this weekend 👋

Tomorrow is the day! Nathan Brake, Senior Machine Learning Engineer at Mozilla.ai, joins GitHub’s Open Source Friday for a live conversation on building AI systems that can adapt as models and providers evolve. Tune in live: gh.io/anyllm

The State of AI report from OpenRouter and a16z captures API-based usage well. But many small models run locally, outside APIs. any-llm-platform helps surface that local activity, adding context the data can’t show. Full breakdown:

An invitation worth sitting with. AI is shaping more of our daily tools than we realize. The choices around agency openness and community will shape what comes next and who it works for.

encoderfile in action 👀 Thanks to Fahd Mirza for publishing a hands-on video that installs and tests encoderfile, showing how transformer encoders can be packaged into standalone executables. Watch here: youtu.be/QMi7Z4kqmBk?...

any-llm gives you one API across LLM providers. > Same code. > Different models. > Easy switching when you need it. The cookbook walks through setup, keys, and generating text across providers. Explore the cookbook: link.mozilla.ai/any-llm-cook...

On Jan 30, Nathan Brake, Senior Machine Learning Engineer at Mozilla.ai, will be joining GitHub’s Open Source Friday. It’s a live conversation about building AI systems that stay adaptable as models and providers evolve, and how open tools help preserve choice. Join us live here: gh.io/anyllm

Say hello to Octonous 🎊 An agent platform that runs real tasks across your apps with clear visibility and approvals. Beta opens in February. Join as an early user: link.mozilla.ai/octonous