On August 18, 2026, the operator of an autonomous Alpaca paper-trading system removed all seven scheduler jobs that directly ran, monitored, reviewed, or published the trading apprentice, leaving 16 j…
A physics researcher with no finance background connected Claude to a US brokerage account via API and ran a one-week live trading experiment with $500, ending at $510.65 (up 2.20%), slightly outperfo…
Alpaca, a developer-first brokerage, has seen monthly active API users grow nearly 4x over the past six months, driven by AI agents, and now manages $1.5 billion in assets under custody. The company r…
A developer running a one-person AI company used Claude Code, an AI coding agent, to build three paper-trading bots for crypto, equities, and options. The agent created working infrastructure but, whe…
Retail traders are increasingly using AI-powered trading bots, once reserved for hedge funds, but risk controls remain largely on the individual trader. A Bloomberg report from July 24, 2026, highligh…
A bug in the evaluation code for GPT-2 style models caused incorrect baseline numbers, but OpenAI's original weights still outperform the author's models on instruction-following tasks. The bug involv…
OpenAI's original GPT-2 small and medium weights consistently outperform custom-trained models in instruction-following evaluations, despite some custom models achieving better test loss, according to…
Bots now generate most web traffic for the first time, with Cloudflare reporting automated traffic crossed 50% in June and now accounts for nearly 58% of all webpage requests. A separate report from H…
AI agents now generate the majority of web traffic, with bot activity surpassing human activity in June, according to CloudFlare, which reported bots accounted for 57.5% of webpage requests. Traffic f…
LMSYS FastChat is the most comprehensive open-source platform for building, serving, and evaluating LLM-powered chatbots in 2026, offering tools for training instruction-tuned models like Vicuna and A…
Alpaca, a developer-focused brokerage infrastructure company, has raised $435 million in a funding round split between $135 million in equity and $300 million in debt to explore prime brokerage servic…
Researchers introduced DominoTree, a training-free best-first draft tree method for speculative decoding that uses Domino's conditional correction to achieve up to 6.6x speedup over autoregressive dec…
Researchers from Meta AI benchmarked the energy costs of inference for LLaMA models on NVIDIA V100 and A100 GPUs across up to 32 GPUs, finding that inference energy consumption is significant and ofte…
Anthropic released a Linux desktop app for Claude Code, offering the same features as macOS and Windows versions. The app requires manual repository setup and only supports Debian/Ubuntu-based distrib…
Public.com integrated its Model Context Protocol server with ChatGPT, enabling users to analyze portfolios, research stocks, and execute trades across stocks, options, crypto, and bonds through natura…
A developer built a local AI research assistant to brief them on investing, after concluding that LLMs are useful for research synthesis, comparative analysis, and thesis stress-testing but not for pr…
A developer trained a large language model from scratch using plain PyTorch, implementing the full post-training pipeline including SFT, reward modeling, DPO, PPO, and GRPO on public datasets, all run…
A journalist let three AIs trade real money in a week-long experiment, testing Robinhood's new agentic trading feature. After testing ~310 strategies, none of the 1× intraday strategies beat QQQ buy-a…
A benchmark called "1rok" that evaluates seven frontier large language models (LLMs) as stock pickers, giving each $100,000 in paper capital and identical tools to select stocks weekly. The project, w…