NewsDeepSeek V4 Expected Feb 17: 1M+ Context, Open-Weight, Consumer GPUWaveSpeed AI·high signalXBlueskyLinkedInCopy linkDeepSeek V4 targets coding dominance with 1M+ tokens, Engram memory, consumer hardware (dual RTX 4090). Open-weight expected. Internal tests claim it beats Claude and GPT on coding.SourceSource pageWaveSpeed AI↳ Follow the threadShared entity / Policy dependencyHacker News Relitigates Anthropic's Ban on Training Models From Your Own Claude OutputsHacker News / Claude Help CenterShared entity / Actor roleAnthropic's Slack 'Claude Tag' Gets ~30% Better at Deciding When Not to Respond, Using Channel Context, Memory and Standing InstructionsReleasebot (Anthropic changelog)Shared entity / Stack layerLaunch HN: Bullet (YC S26) Claims 95.8% on SWE-bench Verified at 119s Per Task — Commenters Call the Benchmark MeaninglessHacker NewsShared entity / Stack layerSamsung's Claude Chip-Verification Rollout: A Month of Work in Two Days, and Unauthorized Edits to Circuit CodeNeowin / Hacker NewsShared entity / ContrastMCP Memory Implements Google's Open Knowledge Format v0.2 Over SQLite FTS5 for Sub-20ms Agent RecallGitHub / Hacker NewsShared entity / Stack layerUnsloth Repositioned From Fine-Tuning Library to Local UI for Running and Training Frontier Open ModelsGitHubShared entity / Stack layerAnthropic's Red Team Gave Three Claude Agents the Same Repo With Conflicting Instructions and Watched Them Write Self-Replicating Sabotage MalwareTechCrunch (corroborated by Unite.AI coverage of the Frontier Red Team post)Shared entity / Stack layerMCP 2026-07-28 Spec Lands: Stateless Core, Governed Extensions, Tasks, and Sandboxed MCP AppsModel Context Protocol Blog