https://zubnet.ai/news/ AI news researched and written by Sarah Chen. Original reporting — no summaries, no press releases. en Tue, 07 Apr 2026 09:31:28 +0000 https://zubnet.ai/sarah.png https://zubnet.ai/news/ https://zubnet.ai/news/githubs-rubber-duck-uses-cross-model-review-catch-ai-coding-errors/ https://zubnet.ai/news/githubs-rubber-duck-uses-cross-model-review-catch-ai-coding-errors/ Different AI families have different blind spots. GitHub's new feature exploits that by using competing models to review each other's code. Tue, 07 Apr 2026 09:30:39 +0000 sarah@zubnet.ai (Sarah Chen) Tools & Dev https://zubnet.ai/news/googles-scion-multi-agent-orchestration-real-isolation/ https://zubnet.ai/news/googles-scion-multi-agent-orchestration-real-isolation/ Google open-sourced Scion, an experimental testbed for running concurrent AI agents in isolated containers with shared workspaces. Tue, 07 Apr 2026 08:05:41 +0000 sarah@zubnet.ai (Sarah Chen) Tools & Dev https://zubnet.ai/news/meta-donates-helion-pytorch-foundation-taking-aim-cudas-kernel-lock-in/ https://zubnet.ai/news/meta-donates-helion-pytorch-foundation-taking-aim-cudas-kernel-lock-in/ The Python DSL promises \ Tue, 07 Apr 2026 07:05:34 +0000 sarah@zubnet.ai (Sarah Chen) Infrastructure https://zubnet.ai/news/pytorch-adds-cutedsl-backend-betting-python-over-c-gpu-kernels/ https://zubnet.ai/news/pytorch-adds-cutedsl-backend-betting-python-over-c-gpu-kernels/ Meta's TorchInductor now supports NVIDIA's CuteDSL as a fourth backend for matrix multiplications, signaling a shift toward Python-based GPU kernel development. Tue, 07 Apr 2026 07:00:52 +0000 sarah@zubnet.ai (Sarah Chen) Infrastructure https://zubnet.ai/news/executorch-joins-pytorch-core-challenging-mobile-ai-deployment-status/ https://zubnet.ai/news/executorch-joins-pytorch-core-challenging-mobile-ai-deployment-status/ Meta's on-device inference runtime becomes official PyTorch project, potentially reshaping how developers deploy AI on phones and edge devices. Tue, 07 Apr 2026 06:35:35 +0000 sarah@zubnet.ai (Sarah Chen) Infrastructure https://zubnet.ai/news/neubird-raises-19m-kill-sre-firefighting-agentic-ai/ https://zubnet.ai/news/neubird-raises-19m-kill-sre-firefighting-agentic-ai/ Another startup promises to solve alert fatigue. But can AI agents really handle production incidents better than humans? Tue, 07 Apr 2026 06:01:07 +0000 sarah@zubnet.ai (Sarah Chen) Industry https://zubnet.ai/news/local-spreadsheet-ai-tool-shows-privacy-first-analysis-actually-works/ https://zubnet.ai/news/local-spreadsheet-ai-tool-shows-privacy-first-analysis-actually-works/ Your Data Assistant processes Excel files locally for $20 lifetime, proving you don't need cloud services for basic AI insights. Tue, 07 Apr 2026 05:50:33 +0000 sarah@zubnet.ai (Sarah Chen) Tools & Dev https://zubnet.ai/news/samsungs-ai-memory-gold-rush-reveals-real-infrastructure-winners/ https://zubnet.ai/news/samsungs-ai-memory-gold-rush-reveals-real-infrastructure-winners/ Record profit forecasts signal memory chips, not flashy models, are where the real AI money flows. Tue, 07 Apr 2026 05:20:32 +0000 sarah@zubnet.ai (Sarah Chen) Infrastructure https://zubnet.ai/news/cursors-warp-decode-claims-18x-gpu-speedupbut-wheres-proof/ https://zubnet.ai/news/cursors-warp-decode-claims-18x-gpu-speedupbut-wheres-proof/ Cursor says their new warp decode technique eliminates MoE overhead on B200 GPUs, but with zero technical details or independent verification. Tue, 07 Apr 2026 05:15:30 +0000 sarah@zubnet.ai (Sarah Chen) Infrastructure https://zubnet.ai/news/metas-eupe-vision-encoder-actually-works-your-phone/ https://zubnet.ai/news/metas-eupe-vision-encoder-actually-works-your-phone/ Meta built a sub-100M parameter vision encoder that handles image understanding, segmentation, and VLM tasks without the usual trade-offs. Tue, 07 Apr 2026 04:45:41 +0000 sarah@zubnet.ai (Sarah Chen) Models & Releases https://zubnet.ai/news/content-moderation-ate-itself-llms-now-moderate-other-llms-end-to-end/ https://zubnet.ai/news/content-moderation-ate-itself-llms-now-moderate-other-llms-end-to-end/ Google research shows LLMs generate training data, detect abuse, handle appeals, and audit their own bias in a closed loop. Tue, 07 Apr 2026 04:05:33 +0000 sarah@zubnet.ai (Sarah Chen) Research https://zubnet.ai/news/neuro-symbolic-ai-promises-100x-energy-cuts-still-research/ https://zubnet.ai/news/neuro-symbolic-ai-promises-100x-energy-cuts-still-research/ Tufts researchers claim their hybrid approach slashes robot AI energy use by 100x while improving accuracy. The catch? It's still proof-of-concept. Tue, 07 Apr 2026 01:05:30 +0000 sarah@zubnet.ai (Sarah Chen) Research https://zubnet.ai/news/metas-open-source-ai-play-faces-reality-check-limited-sources/ https://zubnet.ai/news/metas-open-source-ai-play-faces-reality-check-limited-sources/ Reports claim Meta is planning open-source versions of next-gen models, but sparse details raise questions about the actual scope. Tue, 07 Apr 2026 00:10:31 +0000 sarah@zubnet.ai (Sarah Chen) Models & Releases https://zubnet.ai/news/generalist-claims-99-robot-reliability-wheres-real-proof/ https://zubnet.ai/news/generalist-claims-99-robot-reliability-wheres-real-proof/ GEN-1 robotics model promises production-ready performance on delicate tasks, but the company's claims need serious scrutiny. Mon, 06 Apr 2026 23:30:43 +0000 sarah@zubnet.ai (Sarah Chen) Robotics https://zubnet.ai/news/anthropics-30b-revenue-run-rate-shows-claude-winning-enterprise/ https://zubnet.ai/news/anthropics-30b-revenue-run-rate-shows-claude-winning-enterprise/ Revenue tripled to $30B as Anthropic expands Google partnership for more chips. Enterprise adoption is accelerating faster than expected. Mon, 06 Apr 2026 23:31:51 +0000 sarah@zubnet.ai (Sarah Chen) Industry https://zubnet.ai/news/nvidias-transformer-engine-tutorial-shows-fp8s-real-implementation/ https://zubnet.ai/news/nvidias-transformer-engine-tutorial-shows-fp8s-real-implementation/ A new guide reveals the complexity of actually deploying NVIDIA's mixed-precision training—and why most developers need fallback plans. Mon, 06 Apr 2026 23:25:40 +0000 sarah@zubnet.ai (Sarah Chen) Infrastructure https://zubnet.ai/news/openai-insiders-say-they-dont-trust-altman-control-ais-future/ https://zubnet.ai/news/openai-insiders-say-they-dont-trust-altman-control-ais-future/ The New Yorker interviewed 100+ people about Sam Altman's leadership. Their verdict: he's a people-pleaser with sociopathic tendencies. Mon, 06 Apr 2026 22:01:01 +0000 sarah@zubnet.ai (Sarah Chen) Ethics https://zubnet.ai/news/openai-alumni-launch-100m-fund-while-dodging-conflict-questions/ https://zubnet.ai/news/openai-alumni-launch-100m-fund-while-dodging-conflict-questions/ Zero Shot's timing raises questions about insider access as OpenAI employees cash out through venture investing. Mon, 06 Apr 2026 21:55:34 +0000 sarah@zubnet.ai (Sarah Chen) Industry https://zubnet.ai/news/maine-set-become-first-state-ban-large-data-centers/ https://zubnet.ai/news/maine-set-become-first-state-ban-large-data-centers/ Legislation freezing 20+ MW facilities until 2027 signals growing bipartisan pushback against AI infrastructure boom. Mon, 06 Apr 2026 21:10:33 +0000 sarah@zubnet.ai (Sarah Chen) Policy https://zubnet.ai/news/googles-ios-dictation-app-signals-offline-ai-shift/ https://zubnet.ai/news/googles-ios-dictation-app-signals-offline-ai-shift/ Google quietly launched an offline dictation app using Gemma models, directly challenging Wispr Flow's approach. Mon, 06 Apr 2026 18:55:34 +0000 sarah@zubnet.ai (Sarah Chen) Human-AI Interaction