Results for “context engineering”

…estimates. The design idea—memory as capability, not context—is one of the more practically interesting agent architecture directions in the set. Masked diffusion test-time scaling: @SakanaAILabs shared UnMaskFork, accepted to ICML 2026,…

…Special Shoutout: Swyx will be curating the Data Council AI Engineering Track in Oakland on Apr 22. You can use LATENTSPACE20 for a little discount. {% if medium == 'web' %} Table of Contents [TOC] {% else %} The Table of Contents and…

…priority than where custom instructions are placed in the model context; shared chat examples of prompt engineering and NPC generation. Model Guessing vs Training: One member shared their process by presuming that the model either guesses a…

…and code as an output modality every model bumped up to 128k context length (up from 8k) Training utilized a cumulative of 39.3M GPU hours of computation on H100-80GB (TDP of 700W): 1.5m for 8B, 7m for 70B, 31M for 405B. Llama 3.1 was…

…and Performance: Gemini 2.0 Flash is noted for its larger context window, which some users argue gives it an edge over models like Claude and o1. This capability is seen as crucial in handling real-world software engineering tasks,…

…blogpost goes into more detail on why this is useful: Context: Old BERTS had ~500 token context; ModernBERT has 8k Data: Old BERTS were on older/less data; ModernBERT was trained on 2T, including "a large amount of code" Size: LLMs these…

…interest in the state of the art for vision, especially in the context of building a Retrieval Augmented Generation (RAG) on engineering documents with images and diagrams. Vision SOTA Suggestions: One member touted GPT4v/Geminipro Vision…

…else we dream up together. swyx: Voice AI is the hottest new AI engineering skill! I'll be here - Daily has been in the SF AI Hackathon scene for a very long time and this is the biggest prize set I've seen in a while to learn something…

…@teortaxesTex shared tweets with satirical remarks on AI prompt engineering, tech company behaviors, and AI hype, injecting light-hearted commentary into technical discussions. Casual and Funny Remarks: @richardMCNgo and @teortaxesTex…

…Extensive dialogue on model performance across various contexts, with specific focus on Phi2's U.S. history knowledge, comparison between OpenHermes 2.5 Mistral 7b and Phi2, and experience with LLama 2 chat 7b vs 13b. Highlighted the…

…with third-party services like Cloudflare. Longing for Longer Contexts in Llama 3: Engineers shared their plight over the 8k context window limitation in Llama 3-70B Instruct, pondering over superior alternatives. Suggestions for models…

…GLM-5.3-Flash as a natively multimodal model with a 1M-token context window, 320B total parameters / 18B active parameters, released under the MIT License, and available via weights, API, chat, coding plan, and AutoClaw. Z.ai simultaneously…