Loading activity…
@natolambert
Open model research, founder, director @trillium_labs. Prev. co-led Olmo at Ai2. Writes @interconnectsai, wrote http://posttrainingbook.com
AI
A post argues open-weight cyber models benefit defenders, citing progress finding bugs and the cost of comparable closed models.
AI
Clement Delangue says it omitted the lack of agent containment and monitoring and Hugging Face's open-model defense.
AI
A user calls the result impressive but says the system is multiple times less efficient than Sol.
AI
DeepSeek's README offers offline inference for DeepSeek-V4.1-Flash but lists additional systems needed for a complete service.
Technology
The Wall Street Journal reports the order also bars future purchases of DeepSeek products and services.
AI
A user says trying AI for slides and coherent plans makes them feel AGI will take a while.
AI
A post claims frontier models OpenCode used for defense began refusing help when attack analysis triggered safety guardrails.
AI
A user argues that more compute could accelerate training research, but questions how to measure alignment gains.
AI
A user questions labeling a cyberattack to steal API keys as AI model distillation.
AI
A user contrasts the view that closed models are safe with one emphasizing model-specific risks and messy diffusion.
AI
A user made the remark on October 17 while quoting an OpenCode post.
AI
A post sees Meta gaining momentum and weighs the prospects for US open-weight models to compete with Anthropic and OpenAI.
AI
One commenter says it's hard to separate distillation's contribution from gains made through internal reinforcement learning.
Technology
The company says the model improves benchmarks, task speed and output length over its predecessor.
AI
DeepSeek says the model is live on its website, app and API, and optimized for DeepSeekHarness and Tencent CodeBuddy.
AI
The poster says “Mythos-level” open models appear on track by the end of 2026.
AI
A user jokes that agents discovered a market niche while they slept and would deposit earnings in a USDT wallet.
AI
Reflection says the release includes Mirror CLI, its open-source training harness, and collaborations with 40-plus partners.
AI
A user claims permission would be needed for models above 2T parameters or trained on over 10e26 FLOPS.
AI
One critic cites Muse branding, a revenue gate and restricted-use rules; another compares it unfavorably with Llama 3.1.
