ContentBuffer
A self-taught coder who moved to Silicon Valley at 17 and tested early AI tools like Claude and Midjourney has launched Underdog, an on-device AI assistant. The model, fine-tuned from Qwen3.8 27B, runs entirely on users' devices, making it fast and secure. Underdog can handle everyday tasks and is free, with no ads or data mining. It's backed by Andreessen Horowitz and others. This is a big deal for anyone looking for a privacy-focused AI assistant.
Musubi has launched PolicyLM-1.7B, a decision model designed for real-time content moderation. It applies policies written in plain English to messages in under 50 milliseconds, making it a cost-effective and efficient solution for social platforms. Platform managers can now label content proactively, improving the user experience and reducing the need for human oversight. PolicyLM-1.7B is trained specifically for content moderation and can be customized for different platforms, making it a scalable solution for growing content libraries. Let's dive deeper…
In today's ContentBuffer update:
Lambda Secures $4B for Last Private Round Before 2027 IPO
AI Agents Blocked: Amazon, Walmart, Delta Tighten Rules
Hark Pro: Your Personal AI Assistant for Computer Use
JioHotstar Expands to MENA via Starzplay
5 new AI tools & 5 new AI jobs
More tech news
Latest Development
Image source: techcrunch.com
Summary: Self-taught coder launches Underdog, an on-device AI assistant that uses a 27-billion parameter model. It's invite-only and free, with no ads or data mining. A big win for privacy-conscious users.
Key Points:
Underdog uses a 27-billion parameter reasoning model fine-tuned from Qwen3.8 27B.
The model runs entirely on users' devices, minimizing data transfer between the main chip and graphics chip.
Underdog encrypts keys for email and other accounts users authorize it to access, ensuring security.
The app is free at first and will never be ad-supported, with no user data mining.
Underdog takes a tiny percentage of payment transactions made using Stripe's secure payment rails.
Why it matters: If you're looking for an AI assistant that respects your privacy and doesn't mine your data, Underdog is worth checking out. It uses a 27-billion parameter model that runs entirely on your device, making it fast and secure. The app is free and never shows ads, with no data mining. Backed by Andreessen Horowitz and others, Underdog is a game-changer for privacy-conscious users.
Image source: techcrunch.com
Summary: Musubi's PolicyLM-1.7B applies content policies in real-time, making moderation faster and cheaper. Platform managers can now label content proactively, improving user experience.
Key Points:
PolicyLM-1.7B applies content policies in under 50ms, similar to AI classifiers used by social platforms.
The model outputs binary judgements, simplifying content moderation without the need for complex training.
PolicyLM-1.7B is trained specifically for content moderation, offering a scalable solution for platform managers.
Platform managers can customize the model to fit their specific needs, enhancing content moderation efficiency.
Musubi's interest in decision models dates back to the 2024 GLiNER project, predating TypeSafe AI's Jev.
Why it matters: If you're managing a social platform with growing content, PolicyLM-1.7B offers a scalable solution for real-time moderation. It applies policies in under 50ms, reducing the need for human oversight and improving user experience. Platform managers can customize the model for their specific needs, making it a game-changer for content moderation.
Image source: techcrunch.com
Summary: Lambda secured $4 billion in funding, setting the stage for its 2027 IPO. Investors are betting on Lambda's ability to meet growing demand for GPU capacity, especially with Anthropic's $35 billion commitment.
Key Points:
Lambda's funding round of up to $4 billion is led by Coatue Management and Blackstone.
Lambda's pre-money valuation stands at $14.5 billion, up from its 2025 funding round.
Lambda's backlog grew from $15 billion in June to $50 billion in September, driven by Anthropic's $35 billion commitment.
Lambda's latest $1 billion debt raise comes amid lenders becoming choosier about funding.
Lambda's IPO will join other Nvidia-backed neoclouds like CoreWeave and Nebius in public markets.
Why it matters: Lambda's $4 billion funding round signals stability and growth for developers relying on its GPU capacity. With Anthropic's $35 billion commitment, Lambda's backlog surged, but the high costs of data center buildouts remain a challenge. For teams planning to scale their GPU usage, Lambda's financial stability is crucial.
Image source: techcrunch.com
Summary: Amazon, Walmart, and Delta are blocking AI agents from completing tasks, citing security and user experience concerns. Meta's Muse AI agent is pushing for new standards to facilitate agent-to-agent communication.
Key Points:
Amazon, Walmart, and Delta block AI agents from completing tasks, citing security and user experience concerns.
Walmart partners with Meta's Muse AI agent but still restricts access through its website's security measures.
Industry partners are working on an open standard for AI agent communication, focusing on agent-to-agent communication for online commerce.
Cloudflare offers protection from AI agents and manages automated activity, changing crawler defaults on September 15 to block AI training.
Meta maintains a growing list of 'connectors' (partners) inside the Muse app, focusing on brand partnerships for access.
Why it matters: If you're using an AI agent to automate tasks like flight booking or shopping, you may face restrictions from major retailers and airlines. Walmart's human verification button can fail for AI agents, and Delta is taking steps to protect customers from unauthorized automated activity. This affects the reliability and efficiency of AI agents in real-world applications.
Image source: techcrunch.com
Summary: Hark Pro, a startup less than a year old, launches an AI assistant designed for computer tasks. Free to use, with a subscription tier for heavy users, it promises to revolutionize how you interact with your digital life.
Key Points:
Hark Pro, a startup founded less than a year ago, launches an AI assistant designed for computer tasks.
The service offers a full-screen experience with a central chat dialog and mini-dashboards for various tasks.
Users can configure it to access their digital life, including email, calendar, and hard drive.
The assistant can proactively manage tasks like renewing car registrations and field classic chatbot questions.
The company plans to release an AI-native device in 2027.
Why it matters: If you're tired of manual tasks and want an AI that understands your computer, Hark Pro might be for you. The service is free to use, with a subscription tier for heavy users, and promises to be faster and cheaper at computer-related tasks due to its specialized training. The company plans to release an AI-native device in 2027.
Image source: techcrunch.com
Summary: JioHotstar is expanding into MENA through Starzplay, offering Indian content in multiple languages. The bundled sports plan is priced at AED 49.99.
Key Points:
JioHotstar is expanding into MENA through Starzplay, offering Indian content in multiple languages.
The service will be available through Starzplay's app and website, with a bundled sports plan priced at AED 49.99.
JioHotstar's catalog includes popular shows like 'Bigg Boss' and 'Koffee with Karan'.
The service will offer new originals simultaneously in India and MENA.
JioHotstar's own sports offering in India will not be available in MENA.
Why it matters: JioHotstar's expansion into MENA through Starzplay is a strategic move to tap into the South Asian diaspora communities in the region. The bundled sports plan priced at AED 49.99 offers a competitive option for subscribers looking for Indian content and sports. This move is particularly significant for South Asian diaspora communities in the region, providing them with a familiar content offering.
New Tools & Job
MachineTranslation.com - MachineTranslation.com is an AI-powered translation platform that translates text and documents by comparing outputs from multiple leading AI translation models. Its consensus technology helps users choose the most accurate translation, reducing errors and improving quality across more than 330 languages. What it does Translates text and documents using AI. Compares translations from multiple AI models. Provides a consensus translation to improve accuracy. Supports 330+ languages. Helps users evaluate and refine translations. What makes it good Higher accuracy: Compares multiple AI translations instead of relying on a single model. Better quality: Consensus technology helps identify the strongest translation. Fast and easy: Delivers results in seconds with a simple interface. Broad language support: Covers over 330 languages. Useful for professionals and travelers: Suitable for business, marketing, legal, technical, and everyday communication.
Format Factory - Format Factory is a browser-based workbench for common audio and video conversion jobs. It supports batch processing for video and audio conversion, compression, merging, audio extraction, and audio removal, with transparent credits and temporary results. Try Format Factory for practical media conversion without installer bundles or per-file setup.
HappyHorse.AI - HappyHorse.AI is a multimodal AI video generator for creators, marketers, and social media teams. Users can create videos from text prompts, animate images, or generate videos from reference characters. The platform focuses on browser-based video creation and native audio support.
NeatScribe - NeatScribe is designed for anyone who needs to reuse information from recorded speech. It creates timestamped transcripts, supports translation, and exports results as text files, documents, subtitles, captions, or timed lyrics.
Sakana Marlin - Sakana Marlin is Sakana AI's first commercial product: an autonomous 'Ultra Deep Research' agent positioned as a Virtual Chief Strategy Officer. Given a research topic, Marlin works autonomously for up to roughly eight hours, forming hypotheses, gathering and reconciling information from online sources, and synthesizing a detailed strategy report up to ~100 pages plus executive-summary slides. It builds on Sakana AI research including AB-MCTS (NeurIPS 2025 Spotlight) and The AI Scientist (published in Nature). A closed beta ran from April 2026 with ~300 professionals across finance and consulting.
Agentic AI Engineer - Elastic, the Search AI Company, enables everyone to find the answers they need in real time, using all their data, at scale — unleashing the potential of businesses and people. The Elastic Search AI Platform, used by more than 50% of the Fortune 500, brings together the precision of search and the i
Software Engineer, Guest & Host - Airbnb was born in 2007 when two hosts welcomed three guests to their San Francisco home, and has since grown to over 5 million hosts who have welcomed over 2 billion guest arrivals in almost every country across the globe. Every day, hosts offer unique stays and experiences that make it possible fo
Software Engineer, Staff: Applied AI, Science & Engineering - About Anthropic Anthropic’s mission is to create reliable, interpretable, and steerable AI systems. We want AI to be safe and beneficial for our users and for society as a whole. Our team is a quickly growing group of committed researchers, engineers, policy experts, and business leaders working tog
Staff Applied AI Engineer - Enterprise AI Solutions - About Snorkel Snorkel AI is the frontier AI data lab, helping teams build the data and environments behind high-performing frontier and agentic AI. We combine technology with research-driven AI data development to create datasets, benchmarks, evals, and custom solutions for real-world AI systems. Fo
Data+AI Workforce Transformation Practitioner — Customer Delivery - CSQ327R12 Enterprise customers are investing in AI, but most are stuck between a successful pilot and real organizational adoption. The gap is not technology. It is whether the workforce can change how it works. You will own end-to-end workforce transformation engagements with Databricks' largest cu
QUICK HITS
Mirror Particle Raises $480M to Predict Human Behavior - Mirror Particle raised $480M to build an AI engine predicting human behavior and the reasons behind it. This challenges the current reliance on LLMs for behavior prediction.
Furientis Secures $25M for Missile Production - Furientis secures $25M to produce interceptor missiles for the Pentagon, marking a shift in VC investment towards defense tech.
Founder Summit Offers Discounts for Group Tickets and Bundle Deals - The Founder Summit offers 50% off second tickets and bundle discounts for groups of four or more, making it easier for teams to attend and network.
Anthropic Expands Claude for Startups with Free Year and API Credits - Anthropic expands Claude for Startups with a free year of service, $1k API credits, and more. Great for early-stage companies looking to leverage AI.
LibreOffice Sticks to No-AI Policy, Keeps Data Local - LibreOffice sticks to its no-AI policy, ensuring user data stays local. The Document Foundation vows no AI integration for now, prioritizing data privacy and control.
Wajo's Fo Agent Expands Across Platforms, Grows 10x - Wajo's Fo agent, a privacy-focused personal assistant, now works across iMessage and WhatsApp, and has grown 10x since September. Developers and users concerned about privacy should check it out.
THAT'S A WRAP
Our newsletter reaches a diverse audience of tech professionals, investors, engineers, managers, and business leaders worldwide. Get in touch today.
Be the first to comment
Get the daily edition delivered free every morning.
Musubi introduces PolicyLM-1.7B, a lightweight model for real-time content moderation. It applies policies in under 50ms, making it a game-changer for social platforms. Platform managers can now label content proactively without the need for frequent retraining.
WSL Containers are now generally available, introducing new commands and integrations with Microsoft Defender and Intune. Developers can now build, run, and deploy Linux containers directly on Windows with the wslc CLI.
OpenAI is adding an invisible watermark to ChatGPT and Codex in the EU to comply with the AI Act. Developers can opt-in globally, but it's off by default. The watermark subtly shapes word choices, detectable by a tool.