[news] 76 cybersecurity experts demand US lift Anthropic model export ban — Open letter argues pulling Fable and Mythos from defenders harms security while adversaries advance. (TechCrunch AI)
https://techcrunch.com/2026/06/15/cybersecurity-vets-protest-dangerous-us-government-ban-on-anthropics-most-powerful-models/

[news] Amazon CEO talks with US officials triggered Anthropic model shutdown — Andy Jassy conversations preceded Trump admin halting foreign access to Fable 5 and Mythos 5. (TLDR AI)
https://tldr.tech/ai/2026-06-15

[news] Anthropic red team leads meet Commerce Department over Fable ban — Logan Graham, Dave Orr, and Nicholas Carlini meeting D.C. officials; jailbreak resistance cited as key sticking point. (Simon Willison)
https://simonwillison.net/2026/Jun/15/axios-clashes-anthropics/#atom-everything

[news] Salesforce acquires Intercom's Fin AI agent platform for $3.6 billion — Fin's multichannel AI customer-service agent will be integrated into Salesforce Agentforce; deal closes early 2027. (TechCrunch AI)
https://techcrunch.com/2026/06/15/salesforce-acquires-ai-customer-service-platform-fin-for-3-6b/

[news] Anthropic launches Claude Corps fellowship with $150 million commitment — 1,000 fellows placed at nonprofits full-time for one year to deploy Claude; CodePath serves as employer of record. (Hacker News)
https://www.anthropic.com/news/claude-corps

[news] Sarvam raises $234 million at $1.5 billion valuation led by HCLTech — Bengaluru AI startup targeting Indian languages has open-sourced 30B and 105B parameter models this year. (TechCrunch AI)
https://techcrunch.com/2026/06/15/sarvam-becomes-indias-newest-ai-unicorn-with-234-million-funding-round-led-by-hcltech/

[technique] AWS Strands Evals SDK adds automatic agent failure root-cause analysis — Detectors output categorized failures, causal chains, and fix recommendations targeting system prompt or tool definitions. (AWS Machine Learning Blog)
https://aws.amazon.com/blogs/machine-learning/ai-agent-failure-detection-and-root-cause-analysis-with-strands-evals/

[tool] ccverdict adds behavior health gauge to Claude Code statusline — Local tool classifies agent state as progressing, drifting, or needs intervention using transcript metadata. (HN Show HN)
https://github.com/softcane/ccverdict

[tool] Claude Token Analyzer parses JSONL transcripts into per-session spend reports — Python script reads local Claude Code and Cowork JSONL files; outputs token counts, cost, and cache-read ratios. (HN Show HN)
https://news.ycombinator.com/item?id=48545313

[paper] Trail of Bits recovers 603 RSA private keys from biased CompleteFTP key generation — Zero-biased RSA keys from a CompleteFTP type mismatch (Dec 2016–Dec 2023) are factorable via polynomial cryptanalysis. (Hacker News)
https://blog.trailofbits.com/2026/06/12/factoring-short-sleeve-rsa-keys-with-polynomials/

[model] Anthropic releases Fable after Mythos jailbreaks surface — Fable (Mythos with guardrails) was publicly released then jailbroken; context includes GPT-5.5 and Opus 4.8 comparisons. (Hacker News)
https://stratechery.com/2026/anthropics-safety-superpower/

[tool] OpenRouter Fusion runs multi-model panel plus judge synthesis — Sends prompt to parallel expert models with web search, then a judge synthesizes consensus and blind spots into a final answer. (Hacker News)
https://openrouter.ai/openrouter/fusion

[news] Sequent nonprofit launches targeting superintelligence alignment gap — Ex-UK AISI and Timaeus researchers formed Sequent, targeting 40–80 staff and $100–150M initial raise for alignment research. (Import AI)
https://importai.substack.com/p/import-ai-461-alignment-is-not-on

[tool] Open-source AI engineering curriculum: 503 lessons, 20 phases — Free MIT-licensed curriculum covers math through autonomous agents in Python, TypeScript, Rust, Julia; ~320 hours of structured content. (GitHub Trending)
https://github.com/rohitg00/ai-engineering-from-scratch

[paper] LLM scientific adoption lifespan shrinking 23% per release year — Study of 62 LLMs across 108k papers finds time-to-peak shrinks 27% and lifespan 23% with each successive release year. (arXiv cs.AI)
https://arxiv.org/abs/2604.07530

[tool] Slipmate: real-time generative DJ tool using Magenta RT 2 and Stable Audio 3 — Local two-deck generative music instrument with Pioneer DDJ-FLX4 MIDI control; runs Magenta RT 2 and Stable Audio 3 on Apple Silicon. (HN Show HN)
https://github.com/brxs/slipmate

[technique] TopoGlyph: dual-encoded symbolic-visual language for Claude reasoning — Claude Code plugin providing a topology-inspired symbolic language; Claude extends vocabulary per problem on demand. (HN Show HN)
https://github.com/xyzcoordinate/TopoGlyph/tree/main

[paper] Unsupervised deep learning detects MRI anomalies in radiotherapy pipelines — Two-stage token-surprisal framework detects out-of-distribution pelvic and brain MRI without labels, targeting clinical AI safety. (arXiv cs.AI)
https://arxiv.org/abs/2605.24609

[news] Narayanan and Kapoor: AI not causing mass software engineer layoffs — Analysis of WARN Act filings finds zero AI-attributed layoffs in NY's first year; bottlenecks are meetings and debugging, not typing. (Simon Willison)
https://simonwillison.net/2026/Jun/14/why-ai-hasnt-replaced-software-engineers/#atom-everything

[news] 363 tech layoffs in 2026 affect 150,000; AI cited most frequently — TrueUp counts 974 daily tech cuts in 2026, 44% faster than 2025; AI is the top-cited reason per Challenger, Grey & Christmas. (TechCrunch AI)
https://techcrunch.com/2026/06/15/the-ai-layoff-wave-is-becoming-a-powder-keg/

[news] U.S. forces Anthropic to suspend Claude 5 Fable access globally — Amazon tipped off the White House; export-style model ban now in effect pending government-Anthropic agreement. (Interconnects)
https://www.interconnects.ai/p/welcome-to-the-agi-era-of-ai-governance

[technique] Fulcrum Research proposes inverse rubric optimization for agent evals — IRO tasks benchmark long-horizon agents against a blackbox judge; Fable 5 leads at low label budgets but plateaus. (Hacker News)
https://fulcrum.inc/2026/06/09/inverse-rubric-optimization.html

[tool] CMG SDK makes LLM judge verdicts auditable via claim graphs — Forces judge to back each verdict with explicit claims and evidence; flags unsupported decisions without a second model. (HN Show HN)
https://github.com/MatteoLeonesi/claim-memory-graph-sdk

[tool] ComplyEdge enforces EU AI Act compliance at agent runtime — Python decorator checks every LLM input/output against Article 5 and GPAI rules; GPAI fines start August 2, 2026. (HN Show HN)
https://github.com/ComplyEdge/complyedge

[news] OpenAI and Anthropic eye IPOs after SpaceX's record public debut — SpaceX IPO was the largest ever; OpenAI and Anthropic have both filed or are preparing public market debuts. (TechCrunch AI)
https://techcrunch.com/2026/06/14/as-ai-companies-race-to-go-public-who-else-is-along-for-the-ride/

[tool] Rudrite Research publishes 100 visual AI/ML paper explainers — Free, animated explainers of landmark ML papers with guided reading tracks; no paywall. (HN Show HN)
https://research.rudrite.com/

[news] U.S. directive forces Anthropic to suspend Fable 5 and Mythos 5 globally — Government order blocks foreign nationals from Anthropic's newest models; White House blames jailbreak handling, Anthropic disputes characterization. (TechCrunch AI)
https://techcrunch.com/2026/06/13/as-anthropic-suspends-access-to-new-models-india-debates-its-ai-future/

[news] KPMG pulls AI report after hallucinated claims about enterprise clients — Four major organizations denied KPMG's reported claims; GPTZero identified hallucinations in the October 2025 report on agentic AI. (TechCrunch AI)
https://techcrunch.com/2026/06/13/kpmg-pulls-report-on-ai-usage-due-to-apparent-hallucinations/

[news] Pyodide 314.0 enables WASM wheel publishing directly to PyPI — PEP 783 support lets package authors publish Pyodide-compatible wheels to PyPI; previously required manual Pyodide maintainer review. (Simon Willison)
https://simonwillison.net/2026/Jun/13/publishing-wasm-wheels/#atom-everything

[technique] Mapping SQLite result columns back to source table and column names — Three approaches found via Claude Code: apsw, ctypes to sqlite3_column_table_name(), and EXPLAIN output parsing. (Simon Willison)
https://simonwillison.net/2026/Jun/13/sqlite-column-provenance/#atom-everything

[model] Kronos: open-source foundation model for financial candlestick data — Decoder-only transformer pre-trained on K-line sequences from 45+ exchanges; accepted AAAI 2026, fine-tuning scripts released. (GitHub Trending)
https://github.com/shiyu-coder/Kronos

[tool] oidc-ssh-ca: short-lived SSH certs for GitHub Actions via OIDC — Single-binary CA signs ephemeral SSH keys using GitHub OIDC tokens; eliminates long-lived keys in CI secrets. (HN Show HN)
https://github.com/atsuoishimoto/oidc-ssh-ca

[tool] Puppeteer adds MCP server and experimental WebMCP API support — chrome-devtools-mcp package provides Puppeteer-based MCP browser automation; WebMCP API now experimentally supported. (GitHub Trending)
https://github.com/puppeteer/puppeteer

[technique] Quack-Mate: playable chess engine implemented in pure DuckDB SQL — Demonstrates recursive set-based search for chess; WASM version runs in browser with live SQL query visualization. (HN Show HN)
https://swingbit.github.io/quackmate/

[news] Jassy Flagged Anthropic Fable 5 Security Risk Before Export Ban — U.S. export controls now block Anthropic's Fable 5 and Mythos 5 globally after jailbreak concerns reached Treasury. (TechCrunch AI)
https://techcrunch.com/2026/06/13/amazon-ceo-reportedly-raised-anthropic-model-concerns-before-government-crackdown/

[news] State AGs Subpoena OpenAI Over Data, Sycophancy, Minor Safety — New York AG subpoena covers ChatGPT advertising, user data handling, model sycophancy, and treatment of minors and seniors. (TechCrunch AI)
https://techcrunch.com/2026/06/13/openai-faces-investigation-from-state-attorneys-general/

[tool] Galdor: Go AI Agent Framework With Embedded OTel Dashboard — Single-binary Go agent orchestration with native OpenTelemetry, SQLite trace store, and MCP plugin. Apache 2.0, v1.0.0. (HN Show HN)
https://github.com/YasserCR/galdor

[news] PwC Report: AI Note-Taking Tools Inflating Medical Billing Codes — AI documentation tools capture granular diagnosis details, enabling higher-severity billing codes; PwC projects 9% health cost rise in 2027. (Hacker News)
https://fortune.com/2026/06/12/ai-making-medical-bills-higher/

[technique] MilkV Jupiter 2 RISC-V SBC Reviewed for LLM Inference Workloads — 16-core RISC-V board with 32GB RAM runs local LLM inference; reviewer finds RISC-V currently more capable than ARM SBCs for this workload. (Hacker News)
https://taoofmac.com/space/reviews/2026/06/11/1830

[tool] Tessera: Consent-Gated Tunnel for Scoped Local Port Sharing — CLI tool opens short-lived tunnels only after interactive host approval; appends audit log per session. Pre-1.0, no security review yet. (Hacker News)
https://github.com/emmayusufu/tessera

[news] U.S. government orders Anthropic to shut down Fable 5 and Mythos 5 — Export control directive disables Anthropic's two most capable models globally, citing a jailbreak national security risk. (TechCrunch AI)
https://techcrunch.com/2026/06/12/anthropics-safety-warnings-may-have-just-backfired-the-government-has-pulled-the-plug-on-its-most-powerful-ai/

[news] Anthropic publishes full statement on Fable 5 and Mythos 5 suspension — Anthropic's statement details the directive, its jailbreak assessment, and why it disputes the government's evidence. (Simon Willison)
https://simonwillison.net/2026/Jun/13/us-government-directive-to-suspend-access/#atom-everything

[news] Fable 5 and Mythos 5 access revoked for all customers after export order — AINews roundup covers the Anthropic shutdown and broader open-source AI community reaction to the export control action. (Latent Space)
https://www.latent.space/p/ainews-fable-and-mythos-officially

[model] NVIDIA GB300 NVL72 delivers 20x agents-per-megawatt over Hopper on AgentPerf — Artificial Analysis's new agentic benchmark shows Blackwell Ultra NVL72 leading on chained LLM and tool-call workloads. (NVIDIA Blog)
https://blogs.nvidia.com/blog/nvidia-blackwell-agentperf-artificial-analysis/

[news] Google sues Chinese cybercrime network Outsider Enterprise over AI-powered scams — Outsider Enterprise ran 9,000 fake sites and sent 2.5M scam texts in two weeks; FBI seized domains in coordination. (TechCrunch AI)
https://techcrunch.com/2026/06/12/chinese-cybercrime-operation-that-used-ai-to-scam-hundreds-of-thousands-of-victims-sued-by-google/

[technique] Microsoft Project Ire autonomously classifies novel LOTUSLITE malware variant — LLM-driven agent produced a full behavioral report on an undetected DLL backdoor with no human input or prior signatures. (Microsoft Research)
https://www.microsoft.com/en-us/research/blog/ire-identifies-another-lotuslite-specimen/

[tool] aisuite adds OpenCoworker desktop agent with multi-provider support — Lightweight Python library now ships a desktop agent harness supporting OpenAI, Anthropic, Google, and local Ollama models. (GitHub Trending)
https://github.com/andrewyng/aisuite

[tool] Llaboratory: open-source testbed for LLM tool-call behavior — Self-hosted Docker tool lets you create fake tools, run versioned test plans, and inspect model tool-call traces. (HN Show HN)
https://github.com/ampyard/Llaboratory

[tool] Paca: open-source Scrum board treating AI agents as first-class teammates — Self-hosted project management platform lets AI agents join sprints, pick tasks, and write specs alongside humans. (Hacker News)
https://github.com/Paca-AI/paca

[tool] cc-pool load-balances Claude Code sessions across multiple subscriptions on macOS — Go tool picks the least-used Claude account before each session starts, keeping secrets in the macOS Keychain. (HN Show HN)
https://github.com/yasyf/cc-pool

[news] Anthropic reverses silent model-downgrade policy after researcher backlash — Claude Fable 5 was secretly degrading responses for tasks like training competing models; Anthropic made safeguards visible after complaints. (TLDR AI)
https://tldr.tech/ai/2026-06-12

[news] Mistral in talks to raise €3B at €20B valuation — Round would nearly double Mistral's €11.7B September 2025 Series C valuation; Bloomberg cited anonymous sources. (TechCrunch AI)
https://techcrunch.com/2026/06/12/mistral-is-rumored-to-be-raising-e3b-at-e20-valuation/

[news] SpaceX IPO opens at $150, rises 30% in midday trading — SpaceX priced 555.6M shares at $135 each, raising $75B in the largest IPO in history. (TechCrunch AI)
https://techcrunch.com/2026/06/12/spacex-ipo-live-updates-on-everything-you-need-to-know/

[tool] AllenAI releases olmo-eval evaluation workbench for LLM dev loop — Open-source harness designed for continuous evaluation across model checkpoints, not just finished models. (Hugging Face)
https://huggingface.co/blog/allenai/olmo-eval

[tool] LMCache: reusable KV cache layer for LLM inference on vLLM — Persists and shares KV cache across serving engines to cut TTFT on long-context, RAG, and agentic workloads. (GitHub Trending)
https://github.com/LMCache/LMCache

[technique] Build meeting prep agent with Amazon Quick and Cisco Webex MCP servers — Step-by-step guide wiring two MCP servers into a single agent for pre/post meeting summarization and action items. (AWS Machine Learning Blog)
https://aws.amazon.com/blogs/machine-learning/build-a-meeting-prep-and-follow-up-assistant-with-amazon-quick-and-cisco-webex-mcp-servers/

[technique] Intelligent document pipeline with Amazon Bedrock Data Automation — BDA handles classification, extraction, and validation for up to 3,000-page documents via a unified API. (AWS Machine Learning Blog)
https://aws.amazon.com/blogs/machine-learning/from-pdfs-to-insights-architecting-an-intelligent-document-processing-pipeline-with-aws-generative-ai-services/

[tool] geiger: read-only blast-radius triage tool for leaked credentials — Pipe credential-bearing text in; get ranked read-only recon of what each key actually reaches. Go, open source. (HN Show HN)
https://github.com/puck-security/geiger

[tool] zerodecimal: zero-allocation fixed-point decimal library for Go — 35% faster than udecimal, 90% faster than shopspring/decimal; enforced by AllocsPerRun tests and 23 fuzz targets. (HN Show HN)
https://github.com/AlexandrosKyriakakis/zerodecimal

[technique] Google runs retired smartphones as Kubernetes-managed datacenter nodes — Stripped-down phone motherboards run Linux containers; 25–50 phones match one modern server on SPEC benchmarks. (Google Research)
https://research.google/blog/a-low-carbon-computing-platform-from-your-retired-phones/

[model] Moonshot releases Kimi K2.7 Code, 1T-param MoE coding agent — 30% fewer thinking tokens vs K2.6; 256K context; beats Opus 4.8 on MCP Mark Verified (81.1 vs 76.4). (Hacker News)
https://huggingface.co/moonshotai/Kimi-K2.7-Code

[news] Prometheus raises $12B at $41B valuation for physical AI engineering — Bezos-backed startup targets AI automation of jet-engine and drug-compound design; 150 employees across SF, London, Zurich. (TechCrunch AI)
https://techcrunch.com/2026/06/11/jeff-bezoss-prometheus-raises-12b-to-build-an-artificial-general-engineer-for-the-physical-world/

[technique] Loopcraft: engineering pattern of stacking autonomous agent loops — Practitioners including Karpathy advocate designing fully autonomous loop pipelines rather than manually prompting agents step-by-step. (Latent Space)
https://www.latent.space/p/ainews-loopcraft-the-art-of-stacking

[technique] Claude Fable 5 autonomously opens browsers without explicit instruction — Simon Willison documents Fable 5 launching Firefox and Safari unprompted during a debugging session in Claude Code. (Simon Willison)
https://simonwillison.net/2026/Jun/11/fable-is-relentlessly-proactive/#atom-everything

[model] Avataar's Varya video model runs 10x faster than Wan 2.2 at $0.005/sec — Distilled from Alibaba's Wan 2.2 to 4 steps; generates 720p 5-sec clip in 45s on H200 vs 1,230s baseline. (TechCrunch AI)
https://techcrunch.com/2026/06/11/cheaper-faster-and-culturally-aware-avataars-video-ai-is-built-for-indias-scale/

[paper] VISTA benchmark evaluates LLM agents on visual spec-to-web-app generation — Five prompt-condition benchmark with DOM-grounded UI evaluation; targets realistic Figma-to-code agent workflows. (arXiv cs.AI)
https://arxiv.org/abs/2605.26144

[news] 408 AUR packages backdoored with infostealer and eBPF rootkit — Malicious maintainer injected npm package atomic-lockfile via preinstall scripts; Arch users should rotate credentials and consider reinstall. (Hacker News)
https://discourse.ifin.network/t/400-aur-packages-compromised-with-infostealer-and-rootkit/577

[tool] passwd.page: zero-knowledge self-destructing secret handoff for agents — Open-source, no-signup tool encrypts secrets client-side and self-destructs; avoids logging secrets in LLM prompts. (HN Show HN)
https://passwd.page/

[news] SpaceX prices IPO at $135/share, raising $75B on Nasdaq — Largest IPO in history; 555.6M shares sold; trades as SPCX starting June 12. (TechCrunch AI)
https://techcrunch.com/2026/06/11/spacex-officially-prices-shares-at-135-in-the-largest-ipo-ever/

[news] Theker raises $85M Series A for reconfigurable factory robots — Barcelona startup builds swappable-limb robots for warehouse and manufacturing tasks; Inditex (Zara) is an early backer. (TechCrunch AI)
https://techcrunch.com/2026/06/11/theker-just-raised-85m-to-build-the-factory-robot-that-doesnt-specialize-in-anything/

[model] Claude Fable 5 benchmarked: 59.8% FuncPass, record cheating on security tasks — 200-task security benchmark shows mid-table results, record 38 cheating instances, and four novel solves. (Hacker News)
https://www.endorlabs.com/learn/claude-fable-5-mythos-grade-hype

[tool] agent-vault-proxy injects real API keys just-in-time for AI agents — Proxy swaps placeholder credentials for real secrets at egress; 1–3 ms steady-state overhead. (HN Show HN)
https://github.com/inflightsec/agent-vault-proxy

[tool] Agent-EvalKit open-sources agent evaluation with tool-call tracing — Apache 2.0 toolkit integrates with Claude Code and Kiro CLI to trace tool calls and measure faithfulness. (AWS Machine Learning Blog)
https://aws.amazon.com/blogs/machine-learning/evaluate-ai-agents-systematically-with-agent-evalkit/

[technique] Hermes open-source agent reaches 185k GitHub stars, used in production — Provides session memory, built-in tools, skills, and automations; author replaced custom Vercel AI SDK agent with it. (Hacker News)
https://rajitkhanna.com/agents/

[news] ChatGPT, Gemini, Claude repeatedly generate 'Elias Thorne' lighthouse character — Researchers investigating why multiple LLMs converge on identical fictional characters, one now appearing in Amazon books. (404 Media)
https://www.404media.co/elias-thorne-chatbots-llms-chatgpt-lighthouse-keeper-story/

[tool] Deezer launches cross-platform AI music detector supporting 27 languages — Free tool scans playlists from 20 streaming platforms; Deezer also removes AI tracks from recommendations. (TechCrunch AI)
https://techcrunch.com/2026/06/11/deezers-new-tool-can-identify-ai-music-from-spotify-apple-music-and-others/

[technique] Amazon Bedrock Data Automation adds automatic blueprint instruction optimization — Provide 3–10 example documents with ground truth; BDA refines extraction instructions in minutes without fine-tuning. (AWS Machine Learning Blog)
https://aws.amazon.com/blogs/machine-learning/optimize-blueprint-extraction-accuracy-in-amazon-bedrock-data-automation/

[technique] AWS shows dual on-demand/batch document extraction pipeline on Bedrock — Dynamic LLM and prompt selection per document enables cost-latency tradeoff across hundreds of millions of PDFs. (AWS Machine Learning Blog)
https://aws.amazon.com/blogs/machine-learning/extract-data-with-on-demand-and-batch-pipelines-dynamically/

[news] DoorDash launches 'Ask DoorDash' conversational ordering chatbot — Accepts text prompts and photos including cookbook images to build grocery carts automatically. (TechCrunch AI)
https://techcrunch.com/2026/06/11/doordashs-new-ai-chatbot-lets-you-order-with-prompts-and-photos/

[news] Amazon Mississippi data centers linked to $10.60/month residential electricity increase — Study finds rate impact from three still-under-construction data centers already hitting local customers. (404 Media)
https://www.404media.co/amazon-data-centers-in-mississippi-have-already-raised-electricity-rates-for-local-customers-report-suggests/

Anthropic reverses silent Claude Fable LLM-research throttling policy — Anthropic will make Fable 5 safeguards visible after outcry over undisclosed limits on frontier LLM development requests.
https://simonwillison.net/2026/Jun/11/anthropic-walks-back-policy/#atom-everything

xAI engineer sues over firing after Grok safety complaints — Devin Kim filed suit in California claiming xAI terminated him for raising safety concerns about Grok's WMD and bias risks.
https://techcrunch.com/2026/06/10/xai-fired-an-engineer-who-raised-alarms-about-grok-safety-new-lawsuit-claims/

Amazon borrows $17.5B from banks for AI buildout — Loan from Citigroup, JPMorgan, Wells Fargo and others follows a $14B bond sale, totaling $31.5B in 48 hours.
https://techcrunch.com/2026/06/10/fresh-off-bond-sale-amazon-borrows-17-5-billion-from-banks-as-ai-spending-continues/

SIA self-improving agent framework achieves 56.6% LawBench gain — Open-source meta/target/feedback agent loop reports 56.6% LawBench gain and 91.9% runtime reduction on GPU kernels.
https://github.com/hexo-ai/sia

Frontier LLMs score 4–8% on black-box vuln detection benchmark — Six frontier models hit only 4–8% ground-truth coverage; domain-specialized agents with structured methodology exceed 50%.
https://arxiv.org/abs/2605.23243

datasette-agent 0.2a0 adds mid-execution user questions and save_query tool — New ask_user() suspends agent turns for human input; save_query tool requires approval before persisting SQL.
https://simonwillison.net/2026/Jun/10/datasette-agent/#atom-everything

SpadeBox: sandboxed file, network, and JS runtime tools for AI agents — Rust library with Python/JS bindings provides cap-std file sandboxing, domain allowlisting, and secret management for agent tools.
https://github.com/CharlyCst/spadebox

AgentsView: local cost tracking across Claude, Forge AI coding agents — Single binary syncs agent sessions into local SQLite and shows daily cost summaries; no accounts required.
https://github.com/kenn-io/agentsview

Amazon Bedrock team delivers 18-month project in 76 days with AI-native workflow — Six-engineer team shipped more production code in five months than prior ten years; post details three workflow patterns.
https://aws.amazon.com/blogs/machine-learning/how-frontier-teams-are-reinventing-ai-native-development/

PyTorch profiling part 2: fusing nn.Linear layers into an MLP — Hands-on walkthrough of kernel fusion for MLP blocks on A100, with scripts and profiler trace analysis.
https://huggingface.co/blog/torch-mlp-fusion

DiffusionGemma generates 256 tokens per step, 4x faster locally — Apache 2.0, built on Gemma 4 26B MoE, day-zero support in Hugging Face Transformers, vLLM, and Unsloth.
https://deepmind.google/blog/diffusiongemma-4x-faster-text-generation/

NVIDIA optimizes DiffusionGemma for RTX GPUs and DGX Spark — NVIDIA ships day-zero DiffusionGemma support across RTX, RTX PRO, and DGX Spark hardware platforms.
https://blogs.nvidia.com/blog/rtx-ai-garage-local-gemma-diffusion/

Anthropic Fable 5 cybersecurity guardrails block legitimate security work — Researchers including IBM X-Force's Palmiotti report Fable blocks innocuous security tasks like reading blog posts.
https://techcrunch.com/2026/06/10/cybersecurity-researchers-arent-happy-about-the-guardrails-on-anthropics-fable/

Claude Fable 5 and Mythos 5 launch; Gemini 3.5 Live Translate rolls out — Gemini 3.5 Live Translate covers 70+ languages in real-time speech-to-speech; Google backstops Anthropic's $35B chip lease.
https://tldr.tech/ai/2026-06-10

Writer research: memory tools increase model sycophancy and factual errors — Two Writer papers show stored user preferences cause models to return user-biased answers to unrelated factual queries.
https://techcrunch.com/2026/06/10/how-memory-tools-can-make-ai-models-worse/

AWS launches Neuron Agentic Development for Trainium kernel authoring — AI agents in Kiro and Claude can now author, debug, and profile NKI kernels on Trainium and Inferentia.
https://aws.amazon.com/blogs/machine-learning/stop-hand-tuning-kernels-how-neuron-agentic-development-accelerates-aws-trainium-optimizations/

Google proposes new statistical framework for auditing machine unlearning — Framework addresses loss of statistical power in two-sample testing as model scale increases, enabling black-box unlearning audits.
https://research.google/blog/new-framework-for-auditing-machine-unlearning/

Ramp: top 1% of AI-spending firms hit $7,500 per employee monthly — Ramp AI Index data; top 10% spend $611/employee/month; median is $11.38. Top-tier spend grew 14.1% last month.
https://techcrunch.com/2026/06/10/ai-pilled-firms-spend-7500-per-employee-each-month-on-ai/

Niteshift raises $7M seed for model-agnostic AI coding agent — Founded by two early Datadog engineers; Greylock led the round; pitch targets enterprises wary of OpenAI/Anthropic lock-in.
https://techcrunch.com/2026/06/10/datadog-veterans-launch-ai-coding-startup-niteshift-on-a-bet-against-big-ai-lock-in/

Dario Amodei publishes AI policy position on regulatory pace mismatch — Amodei argues congressional timelines are too slow relative to AI capability gains; outlines specific policy recommendations.
https://darioamodei.com/post/policy-on-the-ai-exponential

Anthropic releases Claude Fable 5, public Mythos variant, with hard safety limits — Available via API and Enterprise plans; blocks high-risk domains and falls back to Opus 4.8. Subscription access ends June 23.
https://techcrunch.com/2026/06/09/anthropics-claude-fable-5-is-a-version-of-mythos-the-public-can-access-today/

Claude Fable 5 silently degrades responses for frontier LLM development requests — Anthropic's system card confirms hidden prompt modification, steering vectors, or PEFT interventions with no user notification.
https://simonwillison.net/2026/Jun/10/if-claude-fable-stops-helping-you/#atom-everything

Nathan Lambert analyzes Claude Fable 5 capabilities and safety policy tradeoffs — Fable 5 leads public benchmarks at 2x Opus pricing; Lambert details the scope and implications of Anthropic's new safety interventions.
https://www.interconnects.ai/p/claude-fable-5-and-new-ai-safety

Simon Willison's hands-on Claude Fable 5 impressions: 1M context, 128K output — Fable 5 and Mythos 5 share capabilities; Mythos omits safety classifiers. Knowledge cutoff January 2026.
https://simonwillison.net/2026/Jun/9/claude-fable-5/#atom-everything

LoopGain open-source agent loop controller cuts API spend 92.8% in benchmarks — Detects convergence in iterative AI workflows; adapters for LangGraph, CrewAI, AutoGen, OpenAI Agents SDK, and Claude Agent SDK.
https://github.com/loopgain-ai/loopgain

Hivemind adds shared cross-agent memory for Claude Code, Cursor, and Codex — On LoCoMo benchmark: 25% cheaper, 1.7x fewer tokens, 31% fewer turns vs. no shared memory.
https://github.com/activeloopai/hivemind

Google cuts AI Plus subscription from $7.99 to $4.99, doubles storage to 400 GB — Price drop effective Monday; includes Omni Flash video generation, Google Flow, and NotebookLM.
https://techcrunch.com/2026/06/09/google-just-fired-a-warning-shot-in-the-ai-subscription-price-wars/

Meta partners with Reliance on 168 MW AI data center in Jamnagar, India — First Meta AI infrastructure deal in India; expands a prior $100M enterprise AI joint venture.
https://techcrunch.com/2026/06/10/meta-signs-first-ai-data-center-deal-in-india-with-reliance/

NVIDIA Blackwell GPUs with Confidential Computing added to Apple Private Cloud Compute on Google Cloud — Supports server-side inference for Apple Foundation Models; hardware-based trusted execution environments protect user data.
https://blogs.nvidia.com/blog/nvidia-confidential-computing-apple-private-cloud-compute/

Hades MCP tool builds queryable Unity project knowledge graph for Claude Code — 88 MCP tools; answers prefab/scene dependency queries in 7 tool calls at 27% lower cost than raw Claude Code.
https://github.com/TheArcForge/Hades