- AI Fire
- Posts
- 🚀 Europe Gets Kolibri
In partnership with
Europe just got a serious open-weight contender. Aleph Alpha’s Kolibri packs 78B parameters, 1M context, and reasoning modes, while activating 3.46B per token.
What's on FIRE 🔥
IN PARTNERSHIP WITH MACPAW
Eney gives you a quick morning briefing, runs ready-made workflows like Start my day, Review my week, or Downloads manager, keeps your desktop and files organized, and lets you point Eney Wand at anything on screen, text, files, images, to get it handled instantly.
AI INSIGHTS
Aleph Alpha just released Kolibri, a new European open-weight reasoning model built for long context, agentic workflows, and regulated enterprise use. Main details:
78.1B total parameters, but only 3.46B active per token thanks to its MoE architecture.
Supports up to 1,048,576 tokens of context, though Aleph Alpha recommends 256K or less for efficient serving.
Includes controllable reasoning modes: none, low, medium, and high.
Supports native tool calling, RAG, coding, and multi-step reasoning.
Trained on nearly 24T tokens, with 21.3% of pre-training data in German.
Built primarily for German and English, with a custom bilingual tokenizer.
Aleph Alpha trained Kolibri on 768 NVIDIA B200 GPUs. It scored 75.5 overall in English and 70.8 in German, beating several similarly sized sparse models despite activating only around 3B parameters per token.
Kolibri is a serious attempt to give Europe a locally built, open-weight alternative for organizations that want strong reasoning and agent capabilities without depending entirely on U.S. frontier labs.
PRESENTED BY MARKETBEAT
10 Stocks at the Center of AI’s $1 Trillion Boom
AI’s growth story is much bigger than chatbots.
The market was already valued at an estimated $800 billion by late 2025, and continued adoption could push it beyond $1 trillion in the years ahead. That growth will require more data centers, computing power, software, networking and intelligent devices.
But the most obvious AI names may not be the only companies positioned to benefit.
MarketBeat’s free 10 Best AI Stocks to Own in 2026 report identifies ten publicly traded companies helping power the next phase of the AI buildout. Some are established leaders. Others occupy less obvious corners of the expanding AI ecosystem.
This free report names 10 stocks positioned to ride that wave before the rest of the market catches on.
ALL YOU NEED TO WIN WITH AI
1. The Easiest Way to Build Your Very Own & Complete Mobile App With AI Before 2027. Claude Code can take you far beyond a basic app prototype. See how to turn an idea into a working mobile project with UI, animations, video, testing, and mobile build preparation.
2. This Claude Code Agentic OS is the New Meta: Voice Control, Skills, and a Second Brain. Something interesting happens when Claude Code gets access to a structured Obsidian vault. This setup may be one of the most interesting Claude workflows to watch right now.
3. These 7 GPT-6 Astra Use Cases Made Us Understand the AGI Hype (Plus 7 Tricks). These 7 examples bring out behaviors we didn’t expect from ChatGPT, and some genuinely feel like a glimpse of where AI agents are heading.
4. This Is How You Can Easily Get a REAL Film Look From Seedance 2.5 (Free Claude Skills). You don’t need a complicated filmmaking workflow to get more cinematic videos. These updated free Claude Skills help you create shots that feel closer to real film.
YOUR $20 AI MARKETING TEAM
OpenAI says agents now produce two-thirds of small-business AI output, up from one-third in April.
Meanwhile, your marketing still runs one prompt at a time. You re-explain your brand voice in every chat, then finish the job yourself. 💔
On Wednesday, Oct 7th, we build a marketing team inside ChatGPT Work, live:
🧠 Build AI teammates for research, content, campaigns and analysis
🛠 Turn repeat tasks into Skills, so you explain them once
⚡️ Hand multi-step work to Agents and get finished work back to review
Which marketing task would you hand to an AI teammate first? Tell us when you sign up. We'll build the top answers live. 👇
TODAY IN AI
AI HIGHLIGHTS
⚡ GPT-6.1 Sol Ultrafast soon? But Claude Opus 5.5 still owns most of the attention right now. Honestly, I’d rather hear what OpenAI is planning for Astra next. Let's see.
🦾 Nobody really knows exactly what work looks like in 5 years ahead, Musk has advice for 20-year-olds trying to stay valuable as AI moves fast. This is worth a read.
💻 DeepSeek just launched Harness v0.2 across all major desktop platforms. It can execute real work & tasks, already supports Plugins and Workspaces out of the box.
🔄 Google is reshuffling Gemini access on Oct. 9: free users move to Flash-Lite only, AI Plus loses Pro, while AI Pro gains Deep Think. Just check your plan until then.
😅 AI is apparently “SI” now, and Musk is wasting zero time. SpaceXAI is being renamed SpaceXSI after Trump pushed “Super Intelligence” as the new term.
🐛 Google just froze its open-source vulnerability rewards program until 2027 after maintainers were flooded with automated, hallucinated and invalid bug reports.
🧠 Reflection hasn’t even launched its model yet, but the setup is: Nvidia backing, billion-dollar compute deals, and a direct target on Chinese open-weight leaders.
💰 Big AI Commitment: Amazon pledged $1B+ over the next five years to support education, job training, and community programs in U.S. regions where it operates data centers.
NEW EMPOWERED AI TOOLS
⚡ CoreSpeed gives AI agents one MCP for apps, memory, and tools, connecting Claude Code, Codex, Cursor, and more.
🧠 Gemini 4 Argon handles complex, long-horizon work with advanced reasoning, coding, and multimodal understanding.
🎥 Smooth Recorder makes Mac screen recordings look professionally edited, with smart zooms, smooth cursors, backgrounds.
📚 Blume 2.0 turns Markdown folders into production-ready docs sites instantly, giving humans and AI agents a fast framework.
AI BREAKTHROUGH
Anthropic is expanding its Anthropic Interviewer, inviting more Claude users to join a roughly 15-minute AI-led voice interview about how AI fits into their lives and what they hope or worry about for the future. Main details:
The interview is conducted by an AI interviewer rather than a traditional survey.
Participants can talk openly about their experiences, concerns, and expectations around AI.
Users may later choose whether their interviews can be made public.
Anthropic previously used the system to interview 1,250 professionals about AI at work.
Anthropic says the format helps collect richer, more nuanced opinions than fixed survey questions.
The earlier research found major differences in how workers experience AI, including excitement about productivity alongside concerns about skill loss, job security, and becoming overly dependent on AI.
Key takeaway: Anthropic is experimenting with AI itself as a research instrument, using conversational interviews to understand how people actually feel about the technology shaping their work and lives.
We read your emails, comments, and poll replies daily
How would you rate today’s newsletter?Your feedback helps us create the best newsletter possible |
Hit reply and say Hello – we'd love to hear from you!
Like what you're reading? Forward it to friends, and they can sign up here.
Cheers,
The AI Fire Team
Reply