Anthropic redeploys Claude Fable 5 with updated classifiers

#1 𝕏

xAI launched Voice Agent Builder, a no-code platform to create human-like voice agents using Grok Voice, available today at $0.05/min.

#2 𝕏

Anthropic is redeploying Claude Fable 5 globally tomorrow with updated classifiers to block more cybersecurity tasks (falling back to Opus 4.8 for coding/debugging).

Also covered by: @Claude

#3 𝕏

Philipp Schmid launched ghealth, an open-source CLI for the Google Health API that pulls 40 Fitbit Air metrics—like per-stage sleep breakdowns (awake, deep, REM) and heart rate—directly in your terminal or into AI agents.

#4 𝕏

NVIDIA AI launched Nemotron-Labs-TwoTower, a diffusion language model that splits a 30B Nemotron-3-Nano-30B-A3B into context and token-generation towers running in parallel using shared pretrained weights. It achieves 2.42× faster text generation while retaining 98.

#5 𝕏

Harrison Chase coins “agentic map reduce” for programmatically spawning subagents to enable deterministic, advanced agent patterns. He points to LangChain DeepAgents’ dynamic subagents docs for a practical implementation.

#6 📝 Mario Zechner

How To Ship Real Code With AI (Not Junk) ft. David Cramer - The Weekly Dev's Brew - David Cramer, co‑founder of Sentry, says LLMs haven't made him faster and rejects the "100× developer" myth, warning mass generation creates junk and security risk—he built a scanner that cost roughly $1,500 to run on a core codebase and found two high‑severity vulnerabilities that more than covered that month's spend. Sentry budgets about $15,000 per developer per year (~$1,500/month) for AI across ~200 engineers, treats it as R&D, and prefers using LLMs to generate deterministic scripts (e.g., parsers when pages change) rather than reprocessing every page with an LLM.

#7 𝕏

Cognition built Agentic MapReduce, a new architecture for whole-codebase reasoning that maps relevant signals across large repos and fans out focused agents over bounded shards.

#8 𝕏

Cognition launched Security Swarm—a new pillar of Devin for Security that bundles tools to find vulnerabilities, validate them at runtime, and automatically ship remediation PRs.

#9 📝 HumanLayer Blog

Announcing general availability for HumanLayer and HumanLayer Cloud→ - HumanLayer and HumanLayer Cloud are now generally available as an AI coding IDE and collaboration platform that the company says helps engineers ship 2–3x faster across the SDLC while maintaining code quality; it supports "bring your own" AI subscriptions (Claude, Codex, etc.) with no separate per-token billing and offers task- and artifact-centric workspaces, versioned artifacts, agent sessions with full visibility, comment-driven design reviews, and local plus cloud daemons that sync via the HumanLayer API.

#10 𝕏

LlamaIndex 🦙 launched Index v2 and the legal-kb reference app, showcasing agentic retrieval that lets AI agents autonomously navigate large document sets using retrieve, read, grep and find APIs.

#11 ▶️

AI Agents are the new SaaS

Greg Isenberg

#12 in

Marc Baselga Anthropic launched Claude Tag, a Slack agent teams can add to specific channels with scoped access to tools, docs, data, and code.

#13 𝕏

Guillermo Rauch launched Vercel’s AI Gateway with trusted inference origins for GLM and recommends trying the high-speed glm-5.2-fast model, served in partnership with @wafer_ai.

#14 𝕏

Jason Zhou is eager to try the Drift Engine. He’s also hand-rolled a task-planning tool that logs an agent’s goal-execution tree to track state, which he finds super helpful.

#15 𝕏

claire vo 🖤 – building @chatprd tested Sonnet 5 and four other LLMs on 62 one-shot SaaS and mobile prototypes, hand-scoring each output. She then compared her ratings to the models’ self-evaluations, showing her taste outperformed theirs.

#16 𝕏

v0 now uses Nano Banana 2 Lite for image generation. Turn on “Generate Images” in the prompt bar to include them in your outputs.

#17 𝕏

Peter Yang announces Claude Fable 5 is back on your Claude subscription through July 7, and drops a no-BS tutorial walking through five hands-on use cases: finding work, life/business advice, making projects ship-ready, planning big ideas, and refactoring your codebase.

Also covered by: @Claude

#18 𝕏

There’s An AI For That shows how usage-based AI products inevitably become banking platforms—complete with balances, holds, credit accounting, and audit trails for every dollar—and lays out seven telltale signs you’re headed that way in a Stigg thread.

#19 𝕏

Peter Yang predicts the AI stack will fragment into multi-model portfolios—low-cost open Chinese models handling routine tasks while frontier models tackle cutting-edge use cases.

#20 𝕏

clem 🤗 – Co-founder & CEO @HuggingFace launched FLARE, a coalition with MIT, Stanford, Princeton, Harvard, Northeastern, Carnegie Mellon and others to introduce a standardized AI flaw‐reporting system that pushes one report to the right developers, safety orgs, and registries...

#21 𝕏

Google AI launched SynthID in 2023 to embed hidden watermarks in AI-generated images, video, audio and text, watermarking over 100 billion images/videos and 60 000 years of audio.

#22 𝕏

Harrison Chase underscores the importance of full visibility into your agent harness and says he’s gravitated toward using LangChain DeepAgents for agents in real-world products.

#23 in

Lenny Rachitsky says PMs must move beyond coordination to AI-native prototyping with real code, conversational MCP data queries, and coding AI agents, and shares Colin Matthews’ mid-2026 framework for harnessing AI to multiply impact.

#24 in

Udi Menkes reports that at AIE swyx framed AI as nested optimization loops powering “intelligent software factories,” while OpenAI’s Codex team showcased their shift from token-maxing to value-maxing chat agents via collaborative feedback loops.

#25 𝕏

xAI launched a turnkey production voice agent platform featuring telephony, knowledge retrieval, developer tools, guardrails, and observability. It supports bring-your-own phone numbers, APIs, and MCPs for seamless integration.