AdvertiseWatchSubscribe
Last scan 55 h agoMonday, November 16, 2026

Early Signal: today's fastest-growing AI repos, models and papers

Edition · Thursday, November 12, 2026114 new signals · 3.5k repos watched · 9 sources
ggml-org/llama.cpp▲ 618stars in the last day 135k total

▲ 618 today

llama.cpp lets anyone run LLMs locally with no dependencies

llama.cpp is a C/C++ library that runs large language models on CPUs, GPUs and even phones without installing heavy frameworks. It supports quantized models and offers an OpenAI-compatible server.

Why it matters. Developers and teams who want fast, cheap inference on their own hardware can spin up a local model without cloud costs.

Try it: curl -LsSf https://llama.app/install.sh | sh

spotted Nov 12

LLMsInference

Today in one paragraph. Today's theme is DIY AI, making everyday gadgets and content yours. From HermesEnabler2, which flips any Apple Watch into a custom Hermes face, to a Chinese GIF Factory that scripts animated stickers with a single prompt, the open-source world is handing the power back to users.

More stories

Reach the people who try new AI tools first. Advertise on Early Signal, from $19 a day →

The repo market

Open-source projects ranked by stars gained today, top 4 of 56

ProjectTodayTotal
vercel-labs/uiNew · 21d oldVercel Labs releases UI SDK for AI agents to build interfacesThe UI SDK bundles keyboard navigation, focus handling, drag-and-drop and ARIA support as reusable behaviors. Agents can combine these primitives to generate accessible UI components across React, Vue, Svelte, Solid and Web Components. ▲ 2013.8k
elementalsouls/Claude-BugHunterClaude-BugHunter turns Claude into a senior red-team assistantClaude-BugHunter is a bundle of 83 Claude Code skills that let the Claude AI act like a bug-hunting researcher or red-team operator. It ships detection patterns, payloads, reporting templates and enterprise attack chains that you trigger with plain English prompts. ▲ 1835.7k
hembramneeraj3-create/aida-64-extremeNew · 10d oldPre-activated AIDA 64 Extreme 2026 ships as one-click Windows installerAIDA 64 Extreme 2026 bundles the full diagnostic, monitoring and benchmarking suite into a single, pre-activated installer for Windows 10/11. The PowerShell script installs everything offline, without needing admin rights or separate licensing. ▲ 1631.3k
AleRibeiro44/archicad-desktop-setupNew · 10d oldArchicad 2026 now installs offline with one-click PowerShell scriptIt provides a pre-activated, self-contained Archicad 2026 installer for Windows 10/11. The PowerShell script extracts, registers, and configures the full BIM suite in a single step, requiring no admin rights or internet after download. ▲ 1621.3k

New models

Trending on Hugging Face, top 2 of 4

From the labs

Straight from the companies building them, top 9 of 23

  1. AWS ML

    Natera launches SWARM, an AI lab-ops assistant on Bedrock

    SWARM is a conversational AI built on Amazon Bedrock AgentCore that stores and retrieves Natera’s internal lab-operations procedures. Lab staff can ask natural-language questions and receive step-by-step guidance without leaving their workflow.

  2. AWS ML

    Four Bedrock knobs that can slash inference costs

    Amazon Bedrock lets you adjust four request-level settings, cache duration, service tier, reasoning effort, and maxTokens, to control how many tokens are processed and which pricing rate applies. Tweaking these options can lower your bill without changing your prompts or application logic.

  3. OpenAI

    Amplitude ships product updates in 2 days with OpenAI

    Wave, Amplitude’s AI-driven product agent, scans usage data, suggests the next feature to build, writes the code, and tracks its impact. The whole loop runs on OpenAI models, letting product teams move from idea to deployment in 1-2 days.

  4. AWS ML

    Cap Amazon Bedrock spend per IAM principal using AWS native tools

    Tag IAM users or roles and let those tags flow into the Cost and Usage Report. Then view Bedrock costs per identity in Cost Explorer and enforce limits automatically with AWS Budgets.

  5. Hugging Face

    IBM releases LLMON markup language with Granite 4.1 models

    LLMON is a lightweight markup that tags parts of a prompt with special tokens so the model can see structure explicitly. IBM has fine-tuned the 3 billion-parameter and 8 billion-parameter Granite 4.1 models to read and produce text wrapped in this notation.

  6. AWS ML

    AWS shows how to cut hidden Bedrock AI costs

    Amazon Bedrock’s blog explains a method to measure the real cost of each AI task, beyond the simple per-token price. It walks through four hidden cost factors and shows how caching, model routing and output limits can lower the bill while proving an agent’s ROI.

  7. AWS ML

    Self-service LLM gateway on Bedrock with OpenAI-compatible API

    LiteLLM runs on AWS Fargate and forwards calls to Amazon Bedrock models through an OpenAI-style endpoint. A single API key gives developers instant access while admins can set per-user costs and quotas.

  8. Google DeepMind

    Google unveils Gemini 4 Hydrogen: cheap, fast AI for daily coding

    Gemini 4 Hydrogen is a large language model tuned for speed and low cost, supporting up to 1 million tokens per request. It targets everyday coding, testing and other agentic workflows and is reachable via Google AI Studio, the Gemini API, and related Google services.

  9. Google AI

    Google Search upgrades to Gemini 4 Hydrogen for visual answers

    Google Search now runs on the Gemini 4 Hydrogen model, delivering more helpful, visual and intuitive answers. The AI can handle complex, multi-part queries, show tables, images and link to relevant sites.

Research

The papers researchers upvoted most today, top 9 of 22

  1. Paper

    New benchmark tests tabular models on generation, imputation, prediction

    One Tab evaluates a single model’s ability to generate new rows, fill missing values, and predict outcomes from the same table. It shows which capabilities overlap and whether a generalist model can handle all three.

    40 upvotes

  2. Paper

    New technique keeps quantized LLM agents on task

    SteerDistill uses a teacher model to steer low-bit LLM agents back toward successful behavior. It fixes the common failure where quantized agents ignore task details or environment cues.

    34 upvotes

  3. Paper

    RL trains LLMs to teach themselves new facts

    The paper introduces a reinforcement-learning loop that rewards a language model for pulling in external information and using it to improve its answers. This lets the model acquire knowledge on-the-fly instead of relying only on its static parameters or simple retrieval.

    28 upvotes

  4. Paper

    SparseRead lets AI agents skim web pages, saving tokens

    SparseRead is a method that lets an AI agent pull only the relevant snippets from a webpage, document, or tool output instead of loading the whole thing. By reading less, the agent uses fewer tokens and answers faster.

    19 upvotes

  5. Paper

    New routing method mixes LLMs per request to cut costs

    It lets a service pick a custom mix of language models for each prompt, adjusting the number and size of models to hit a target quality or budget. This replaces the usual one-model-per-query approach and can lower compute spend while keeping answers acceptable.

    14 upvotes

  6. Paper

    GTAGym brings AAA game tasks to RL research

    GTAGym is an open-source environment that lets reinforcement-learning agents practice and be evaluated on real-world tasks inside commercial AAA video games. It provides a gym-compatible API and automatic scoring so you no longer need human annotators to judge success.

    13 upvotes

  7. Paper

    New Interactive Visual Gym lets AI agents train in simulated tasks

    Interactive Visual Gym is a benchmark suite that creates virtual environments with simulated users to evaluate visual interaction models in closed-loop settings. It shows how an assistant’s responses influence later user actions and overall task success.

    13 upvotes

  8. Paper

    New serving system adapts LLM compute for agent workloads

    The paper proposes a serving architecture that lets large language model agents dynamically adjust compute allocation and KV-cache handling at runtime. This flexibility aims to improve latency and resource utilization for ever-changing agent tasks.

    13 upvotes

  9. Paper

    New study questions if neural nets truly share representation space

    The paper re-examines the Platonic Representation Hypothesis, which says larger models develop similar internal representations across tasks. It presents recent findings that this alignment may not keep improving as models scale.

    7 upvotes

The conversation

What builders are arguing about on Hacker News, top 8 of 9

  1. Hacker News

    Rust-powered embroidery editor lets AI design stitches from SVG

    ZigZag is a desktop app that turns vector graphics into embroidery files and can talk to AI agents via its Model Context Protocol. It exports common formats and can stream designs straight to a Brother Skitch over Bluetooth.

    719 points54 commentsRead the thread

  2. Hacker News

    Google rolls out Gemini 4 Hydrogen, cheap fast AI for coding

    Gemini 4 Hydrogen is a large language model optimized for speed and low cost, aimed at everyday coding and agentic tasks. It offers a 1 million-token context window and is priced at $1.50 per million input tokens and $7.50 per million output tokens.

    558 points108 commentsRead the thread

  3. Hacker News

    Google adds Orchestrator to toggle Gemini 4 Hydrogen and Carbon

    Orchestrator is a new model selector in Google Antigravity that dynamically switches between the cheap Gemini 4 Hydrogen model and the higher-quality Gemini 4 Carbon model as a task evolves. This keeps simple steps fast and inexpensive while calling the stronger model only when needed.

    495 points47 commentsRead the thread

  4. Hacker News

    LLMs now run on old TV boxes and decade-old iPads

    smolmaxxing finds the biggest language model a low-end device can actually run and then pushes it further with custom kernels, smarter memory handling and weight streaming. It demonstrates a 0.8 B-parameter model at 16 tok/s on a cheap TV box and a 1.7 B model at 25 tok/s on a 2013 iPad.

    465 points69 commentsRead the thread

  5. Hacker News

    Apple speeds up vector search for RAG

    Apple’s new method speeds up maximum inner product search, the core operation behind recommendation engines and retrieval-augmented generation. It reduces the need for exhaustive comparison by building an index that accounts for the difference between short queries and long document keys.

    239 points29 commentsRead the thread

  6. Hacker News

    Retirement Funds Are Bankrolling the AI Data-Center Boom

    The article shows that U.S. workers’ 401(k)s and IRAs are being invested in companies building massive AI data centers. Even as many Americans protest these projects, their retirement money is quietly financing them.

    165 points186 commentsRead the thread

  7. Hacker News

    Browser CAD gets AI teammate for real-time drafting

    Aircad runs 2D drafting and 3D solid modeling entirely in the browser. It adds multiplayer editing and an on-screen AI that can follow plain-language commands to draw, model, and render.

    92 points38 commentsRead the thread

  8. Hacker News

    Klad lets AI agents paint alongside you in real time

    Klad is a 2-D painting and animation editor you share with a coding agent. Both you and the agent can draw using the same brushes, layers and timeline, and each stroke appears instantly for the other.

    83 points32 commentsRead the thread

Get the signal first.

Join the list for the morning email: the day's biggest AI launches, explained in five minutes. Unsubscribe any time.

Prefer a reader? Follow the RSS feed.