Results for “context engineering”

…supporting multimodal tasks. All models support a 128K token context length. Performance: The 1B and 3B models outperform Gemma 2 2.6B and Phi 3.5-mini on key tasks, while the 11B and 90B vision models are competitive with Claude 3 Haiku…

…model). Infra mismatch: serverless vs agent loops: A pointed engineering claim is that agentic systems need persistent, long-running execution and state management; “serverless” patterns force brittle networking workarounds for multi-step…

…3 Opus, best of 4 runs. We are working on clustering and flow engineering with Haiku. OpenAI and Google AI Announcements OpenAI GPT-4o Launch: @sama noted the aesthetic difference between OpenAI and Google's AI announcements. @zacharynado…

…(design story here) a new mascot, Clawd Claude API: a new context editing feature and memory tool to the Claude API that lets agents run even longer and handle even greater complexity. Renaming the Claude Code SDK to Claude Agent SDK.…

…3 Opus, best of 4 runs. We are working on clustering and flow engineering with Haiku. Claude 3.5 Sonnet from Anthropic Impressive performance: Claude 3.5 Sonnet secured the #1 spot in Coding Arena, Hard Prompts Arena, and #2 Overall,…

…the latest image generation model from OpenAI that supports contextual image generation and editing. vLLM for RLHF: @vllm_project highlighted OpenRLHF, a framework to use vLLM for RLHF, driving many design and implementation of vLLM's…

…succinctly matches a broader shift toward harness-centric engineering @fchollet. Related advice includes using asserts heavily in Python/ML code to fail fast @gabriberton, building both end-to-end and incremental evals for long-running…

…3 Opus, best of 4 runs. We are working on clustering and flow engineering with Haiku. Meta Llama 3 Release Model Details: @AIatMeta released Llama 3 models in 8B and 70B sizes, with a 400B+ model still in training. Llama 3 uses a 128K vocab…

…Thread and resources from @stuart_sul, part 2/links, with context from @simran_s_arora. Inference at scale hiring (OpenAI): @gdb outlines OpenAI’s focus areas: forward-pass understanding/optimization, speculative decoding, KV offloading,…

…Releases: Embeddings, Image/Video and Open ModelsPerplexity contextual embeddings: pplx-embed-v2-context-9b-preview is open on Hugging Face (@perplexity_ai). Method: The model encodes the whole document once and pools chunk vectors…

…for Gemma 3 with potential multimodality and longer context. Mistral AI is commended for its rapid progress despite limited resources compared to larger companies. Users suggest normalizing comparisons based on team size and available…

…presumably due to unexpectedly high usage. The Rise of "Context Engineering": The term "prompt engineering" is being challenged, with prominent voices like @jd_pressman and @nptacek advocating for "context engineering" as a more accurate…