Bullet, a coding agent launched from YC's Summer 2026 batch, claims 95.8% on SWE-bench Verified at 119 seconds per task, but its founders concede the benchmark is saturated and the speed pitch is the …
Mixedbread shipped Toast 1, a specialized search agent that matches Claude Opus 5 and GPT-5.6 Sol on deep-search benchmarks while running up to 10× cheaper and 12× faster, according to vendor-run test…
Anthropic published a guide titled "Maximizing the value of your Claude Code sessions" that reveals how its prompt caching billing mechanics affect developer workflows, with tips like using /compact b…
Tencent's Hunyuan3D team published WorldClaw, a system that turns text prompts into editable 3D worlds, but its core orchestration relies on Anthropic's Claude Opus 4.8 and OpenAI's GPT-Image-2, with …
OpenAI's first telemetry-based study, 'How Organizations Use AI: Evidence from ChatGPT,' analyzing 17 million messages from over 1,500 organizations, finds enterprise AI adoption is broad but shallow,…
Only 13% of games released on Steam in 2024 ran on custom engines, down from 71% in 2012, yet those games pulled in roughly 40% of Steam's revenue, according to Video Game Insights. Team KwaKwa, the L…
Anthropic announced that future Claude models will embed an invisible, cryptographically verifiable watermark in all generated text across the API, Claude Code, and consumer apps, with a detection API…
University of Colorado Boulder professor Tom Yeh's AI by Hand project, which teaches transformer mechanics through pencil-and-paper worksheets and an MIT-licensed Excel repository with about 6,000 Git…
Alibaba's Qwen team released two open-weight models this week, with the 2.4-trillion-parameter Qwen3.8-Max grabbing headlines but the Apache-licensed, multimodal Qwen3.8-27B being the practical choice…
Google's HEIR compiler toolchain for fully homomorphic encryption has reached four working demos of AI inference on encrypted data, including a deep-learning recommendation model, credit-card fraud de…
Alibaba's Qwen team released Qwen3.8-27B, a dense, natively multimodal 27B model under Apache 2.0 with weights on Hugging Face, designed for single-GPU agentic work. The model features a 3:1 hybrid of…
OpenAI and Cerebras are previewing an "Ultrafast" mode for GPT-5.6 Sol that delivers 750 tokens per second, roughly 14x the standard tier's ~53 tokens per second, marking the first time a closed front…
Four independent Dev.to authors published nearly identical Claude Code workflows within the past few months, each describing the same context-plan-delegate-verify loop, a convergence that reflects the…
Z.ai shipped GLM-5.3 on August 14 with Terminal-Bench 3.0 scores jumping from 4.6% to 28.3% and DeepSWE v1.1 from 46.2% to 66.9% on the same 743B base model as GLM-5.2, attributing all gains to scaled…
A solo developer's $2,430 refactor of 189 files in a 717,725-line TypeScript codebase probes agentic coding's blind spot: changes with no test oracle. Joël Abenhaïm, a Paris-based developer, used AICo…
RAGFlow, an open-source RAG engine launched in April 2024 by InfiniFlow, has reached 88,000 GitHub stars and continues trending in August 2026, despite repeated declarations that RAG is dead. The proj…
Google shipped Gemini 3.7 Flash on August 13, three weeks after Gemini 3.6 Flash, with benchmark gains in agentic coding tasks such as DeepSWE v1.1 jumping from 49.0% to 65.3%. The launch price is $0.…
Anthropic and alignment researchers including Caspar Oesterheld and Emery Cooper released the Conceptual Reasoning Index (CRI), a 0–100 benchmark scoring AI reasoning in domains without ground truth, …
OpenAI shipped GPT-5.6 on July 9 with three tiers—Sol, Terra, and Luna—priced at $5/$30, $2.50/$15, and $1/$6 per million input/output tokens, respectively, marking a shift to a tiered product structu…
DeepSeek released DeepSeek Harness v0.1, an MIT-licensed, Node.js-based agent runtime now in developer preview on GitHub, where it surpassed 30,000 stars within hours. The harness makes the agent loop…