Together AI: Deploy Any HuggingFace Model in a Single Session
Together AI's AI Native Cloud now deploys any Hugging Face model in a single session, eliminating the multi-day setup gap for custom model inference.
Together AI's AI Native Cloud now deploys any Hugging Face model in a single session, eliminating the multi-day setup gap for custom model inference.
Mirage by Strukto: unified virtual filesystem mounts S3, GitHub, Gmail, Notion and more under Unix semantics for AI agents.
LlamaIndex launched LlamaParse Mobile — an iOS/Android app that converts a phone camera photo into clean, copyable document text in under one minute.
LangChain DeepAgents Harness Profiles deliver 10–20pt tau2-bench gains via per-model system prompt and middleware overrides; harness is now a first-class versioned object.
deepagents ships ACP built-in: any interface (CLI, TUI, GUI, IDE) drives the same agent harness without lock-in. toad TUI and JetBrains IDE integration available on day one.
DS2API exposes DeepSeek Web as OpenAI/Claude/Gemini wire-compatible APIs — reverse-engineered, disclaimer-heavy, but a clear signal of demand for free API access to frontier-adjacent models.
GitHub Next's ACE enters technical preview: Slack-style multiplayer coding sessions, per-session microVMs, shared terminals, and collaborative plan docs — alignment as the new bottleneck.
Memori hits 81.95% LoCoMo accuracy at just 1,294 tokens/query — 67% smaller prompts than Zep, 20x cheaper than full-context — with MCP server and multi-agent attribution model.
memsearch by Zilliz: one memory backend across Claude Code, Codex, OpenClaw, and OpenCode — markdown SSOT, local ONNX embeddings, BM25+dense+RRF hybrid search.
DeepAgents middleware gains adoption with batteries-included defaults and full hook-based customisation, now running in production for manufacturing diagnostics.
LangSmith Fleet ships file editing (images, PDFs, text) and a presentation renderer — agents write slides that render live inside the app via slash commands.
Anthropic's Claude Cowork now supports interactive charts and diagrams in beta across all paid plans, expanding the collaborative workspace.
LlamaIndex open-sources LiteParse—a VLM-free, ML-free PDF parser using a 6-step grid projection algorithm that handles tables and complex layouts fast.
Curated AI insights — sent when there's something worth your inbox.