Boston Children's Diagnoses 40+ Rare Diseases with AIVercel Switches to Per-Unit Pricing for Function CallsLangChain Launches LangSmith PlatformAnthropic Raises $965B for Opus 4.8 LaunchAnthropic hits $47B annual revenue run-rateVercel Warns Against AI Inference TheftRL Prevents Forgetting Better Than SFT in LLMsVFEAgent automates finite element analysisLLM agents overcome phenotype annotation bottleneckOrthogonal concept erasure for diffusion modelsEfficient AR to diffusion model adaptationAgent tool planning for molecular lead optimization9 demos of Gemini Omni and 3.5 capabilitiesGoogle Creates I/O 2026 Quiz via AI StudioOpenAI launches Rosalind BiodefenseWaterloo students showcase AI education prototypesBraintrust speeds up coding with CodexClaude Code Adds Automatic Plugin LoadingOpenAI Codex Enhances Environment DiagnosticsDatasette 1.0a31 Adds Write Query SupportBoston Children's Diagnoses 40+ Rare Diseases with AI

Vercel is changing function billing from package-based to per-unit pricing for Pro and Enterprise customers. Existing rates stay until the current billing cycle ends, after which costs align directly with usage.

LangChain released LangSmith Platform offering real-time agent monitoring, performance evaluation, and deployment tools. Developers can track agent behavior and improve model performance.

Anthropic secured $965B in Series H funding and released Claude Opus 4.8 with Dynamic Workflows and ultracode features. The new model achieves 92% code generation accuracy.

Anthropic raised $6.5B in Series H funding, revealing its annual revenue run-rate reached $47B earlier this month. Enterprise adoption has continued growing since their $4.5B Series G in February, valuing the company at $65B.

Vercel warns that frontier model inference costs ($2/prompt) are 1Mx higher than HTTP requests, making AI theft lucrative. Recommends access controls and rate limiting.

New research reveals RL preserves LLM capabilities better than SFT by maintaining specific neural circuits, providing insights for preventing catastrophic forgetting during training.

VFEAgent is a multimodal agent framework enabling end-to-end automated Finite Element Analysis. It addresses the complexity of traditional FEA workflows and reduces dependency on domain expertise.

Frontier LLM-based agents solve the ontology curation bottleneck for phenotype annotation, automating the labor-intensive process of linking free-text phenotype descriptions to ontology terms.

Orthogonal concept erasure removes unwanted content from diffusion models without retraining, solving computational cost issues of existing methods for content moderation.

A new method adapts Autoregressive models to diffusion models using strictly causal and elastic horizons, solving structural incompatibility while enabling efficient parallel text generation.

Molecular lead optimization is accelerated through agentic tool planning that automates multi-stage optimization, improving ADMET properties to transform early compounds into drug candidates.

Google released 9 videos demonstrating real-world applications of Gemini Omni and 3.5, including complex reasoning, multimodal understanding, and long document processing. The demos were showcased at Google I/O 2026.

Google used AI Studio to build an interactive quiz showcasing key I/O 2026 announcements, demonstrating the latest AI capabilities through vibe coding techniques.

OpenAI expands access to GPT-Rosalind for vetted developers and U.S. government partners in biodefense, public health, and pandemic preparedness through Rosalind Biodefense.

University of Waterloo students developed AI prototypes including sign language tutors, reshaping the future of education and work through practical AI applications.

Braintrust engineers leverage GPT-5.5’s Codex to accelerate experiments and code generation. The tool automates test case creation, result analysis, and converts natural language requirements into executable code.

Claude Code v2.1.157 adds automatic plugin loading from .claude/skills, claude plugin init command, and autocomplete for plugin arguments streamlining development.

OpenAI Codex 0.135.0 adds codex doctor for rich environment diagnostics, improved remote connection details, and enhanced Vim mode for better debugging.

Datasette 1.0a31 introduces write query execution and stored query saving (renamed from ‘canned queries’) for authorized users, significantly improving database interaction capabilities.