A developer built Racksmith, an autonomous modular synthesizer planner that pairs a Gemini 3.8 Flash agent via the Vercel AI SDK with a Sanity Knowledge Lake through a Model Context Protocol (MCP) ser…
A developer who left Google after nine years built a personal AI financial advisor system that consolidated roughly a dozen brokerage, retirement, crypto and credit accounts into a single balance shee…
Goodhart Labs released HoneyBench v0.1, a nine-task benchmark for reward hacking in frontier models including Opus 5.5, Fable 5.1, GPT-6 Astra, Gemini 3.8 Flash, Grok 4.7, and DeepSeek V4 Pro. The lab…
Google's Gemini 4 Argon scored 53 on the Artificial Analysis Intelligence Index, tying OpenAI's flagship for second place among AI labs behind Anthropic's top models, according to Artificial Analysis.…
Google's free Antigravity tier exhausted its weekly quota on every model it serves — Gemini 3.8 Flash, Claude Opus 4.6 and GPT-OSS — in a single afternoon of testing, with each pool reporting "Individ…
A developer replaced the Rust `clap` crate with LLM-generated argument-parsing code written by Gemini 3.8 Flash via OpenRouter, cutting the compiled executable size by 300 kB and argument parsing time…
Anthropic's Claude Opus 5.5, which shipped this week, leads SimpleBench at 88.4% and ranks as Anthropic's best vision model to date, according to AI Battle and @skalskip92, performing better than Fabl…
Arize AI's Jev 1.13.0 classifier matched LLM judges on accuracy while its returned probability strongly predicted its own correctness across 36,190 judgments, according to an Arize benchmark spanning …
Adding the instruction "Use null for any field whose value is not on the page. Do not guess." cut invented fields from 70.7% (405 of 573 missing fields) to 20.2% (116 of 574) across 16 models tested b…
Google released Gemini 3.8 Flash and the vulnerability-focused Gemini 3.8 Flash Cyber on September 2, 2026, with introductory pricing of $0.75/$3.75 per million input/output tokens that doubles to $1.…
A developer built "Blog vs Bytecode," a Kaggle benchmark of 28 data-science code snippets paired with blog-style claims, half of which hide real methodological errors such as scaler leakage or shuffle…
GPT-6 Astra and Claude Opus 5.5 lead the Terminal-Bench-Science benchmark by roughly 20 points over Fable 5.1, with the strongest model outside those two labs, Qwen3.8 Max, at just 12%, according to t…
Gemini 3.8 Flash is suffering lag and repeated retryable error loops in Google's AI Studio, according to a user report on the free plan, forcing developers to roll back projects to earlier versions. T…
A developer spent three hours at Google's Build with Gemini Sunnyvale event directing Antigravity, Google's agent-first coding tool, which built, deployed, published, and self-recorded a demo video of…
A developer-run head-to-head test found that Gemini 3.8 Flash at high thinking effort matched Claude Opus 4.6 on machine-checked instruction following, both scoring 17 of 17 on a seven-constraint arti…
Specific Labs' Real-SWE benchmark, which evaluates frontier models on private enterprise codebases across 640 rollouts, found that Fable 5.1 running in Claude Code achieved the highest first-try resol…
Google Cloud software engineer Ryan Lopopolo, who coined the term "agent harness," described running fully autonomous coding workflows in production on The Agent Factory podcast, saying he has not ope…
Anthropic's life sciences lab found a novel enzyme system with CRISPR-like repeated DNA sequences in a jumbo phage after giving Claude only a high-level prompt, a result CRISPR pioneer Feng Zhang call…
Developer Angelotc released antigravity-seo, a Go-based SEO and Generative Engine Optimization suite for Google Antigravity that replaces Python/Playwright runtimes with a single Go binary called seo-…
A developer's controlled comparison of two candidate planner models for an agentic publishing pipeline found Gemini 3.8 Flash (High) and Claude Opus 4.6 (Thinking) both scored 17 of 17 on machine-chec…