https://learn.physea.ai/https://learn.physea.ai/about/https://learn.physea.ai/agents-and-tooling/https://learn.physea.ai/agents-and-tooling/ai-agents/chatbot-or-agent/https://learn.physea.ai/agents-and-tooling/ai-agents/the-agent-loop/https://learn.physea.ai/agents-and-tooling/ai-agents/the-three-ingredients/https://learn.physea.ai/agents-and-tooling/ai-agents/what-agents-are-good-at/https://learn.physea.ai/agents-and-tooling/ai-agents/workflows-vs-agents/https://learn.physea.ai/agents-and-tooling/ai-harnesses/anatomy-of-a-harness/https://learn.physea.ai/agents-and-tooling/ai-harnesses/autonomy-ladder/https://learn.physea.ai/agents-and-tooling/ai-harnesses/coding-harnesses-compared/https://learn.physea.ai/agents-and-tooling/ai-harnesses/form-factors/https://learn.physea.ai/agents-and-tooling/ai-stacks/eval-and-observability-stack/https://learn.physea.ai/agents-and-tooling/ai-stacks/the-agent-stack/https://learn.physea.ai/agents-and-tooling/ai-stacks/the-fine-tuning-stack/https://learn.physea.ai/agents-and-tooling/ai-stacks/the-local-inference-stack/https://learn.physea.ai/agents-and-tooling/ai-stacks/the-rag-stack/https://learn.physea.ai/agents-and-tooling/ai-stacks/what-is-an-ai-stack/https://learn.physea.ai/agents-and-tooling/mcp-and-tool-use/authorization/https://learn.physea.ai/agents-and-tooling/mcp-and-tool-use/how-mcp-is-built/https://learn.physea.ai/agents-and-tooling/mcp-and-tool-use/the-problem-mcp-solves/https://learn.physea.ai/agents-and-tooling/mcp-and-tool-use/tool-use/https://learn.physea.ai/agents-and-tooling/rag-retrieval/embeddings-and-similarity/https://learn.physea.ai/agents-and-tooling/rag-retrieval/grounding-and-where-it-breaks/https://learn.physea.ai/agents-and-tooling/rag-retrieval/two-phases/https://learn.physea.ai/agents-and-tooling/rag-retrieval/vector-databases/https://learn.physea.ai/agents-and-tooling/rules/decision-vs-action/https://learn.physea.ai/agents-and-tooling/rules/least-privilege/https://learn.physea.ai/agents-and-tooling/rules/runtime-guardrails/https://learn.physea.ai/agents-and-tooling/rules/system-prompt-vs-rules/https://learn.physea.ai/agents-and-tooling/skills/an-open-standard/https://learn.physea.ai/agents-and-tooling/skills/progressive-disclosure/https://learn.physea.ai/agents-and-tooling/skills/skill-tool-or-prompt/https://learn.physea.ai/agents-and-tooling/skills/what-is-inside-a-skill/https://learn.physea.ai/agents-and-tooling/workflows/durable-execution/https://learn.physea.ai/agents-and-tooling/workflows/orchestration-patterns/https://learn.physea.ai/agents-and-tooling/workflows/structural-fixes/https://learn.physea.ai/agents-and-tooling/workflows/what-is-agent-orchestration/https://learn.physea.ai/agents-and-tooling/workflows/why-reliability-matters/https://learn.physea.ai/companies/https://learn.physea.ai/foundations/https://learn.physea.ai/foundations/capabilities-and-limits/hallucination/https://learn.physea.ai/foundations/capabilities-and-limits/no-real-time-knowledge/https://learn.physea.ai/foundations/capabilities-and-limits/understanding-and-counting/https://learn.physea.ai/foundations/capabilities-and-limits/what-ai-does-well/https://learn.physea.ai/foundations/key-terms-glossary/embedding-and-quick-reference/https://learn.physea.ai/foundations/key-terms-glossary/inference-and-fine-tuning/https://learn.physea.ai/foundations/key-terms-glossary/model-parameters-training/https://learn.physea.ai/foundations/key-terms-glossary/prompt-token-context/https://learn.physea.ai/foundations/tokens-and-tokenization/how-text-becomes-tokens/https://learn.physea.ai/foundations/tokens-and-tokenization/tokens-cost-and-context/https://learn.physea.ai/foundations/tokens-and-tokenization/what-is-a-token/https://learn.physea.ai/foundations/tokens-and-tokenization/why-tokens-not-words/https://learn.physea.ai/foundations/training-vs-inference/fine-tuning/https://learn.physea.ai/foundations/training-vs-inference/learning-at-inference/https://learn.physea.ai/foundations/training-vs-inference/pretraining/https://learn.physea.ai/foundations/training-vs-inference/two-modes/https://learn.physea.ai/foundations/what-is-a-neural-network/how-it-learns/https://learn.physea.ai/foundations/what-is-a-neural-network/the-brain-analogy/https://learn.physea.ai/foundations/what-is-a-neural-network/the-parts/https://learn.physea.ai/foundations/what-is-ai-ml-llm/how-they-nest/https://learn.physea.ai/foundations/what-is-ai-ml-llm/what-makes-an-llm/https://learn.physea.ai/foundations/what-is-ai-ml-llm/what-the-words-mean/https://learn.physea.ai/history-and-context/https://learn.physea.ai/history-and-context/deep-learning-revolution/backpropagation-at-scale/https://learn.physea.ai/history-and-context/deep-learning-revolution/big-data-and-imagenet/https://learn.physea.ai/history-and-context/deep-learning-revolution/the-alexnet-moment/https://learn.physea.ai/history-and-context/deep-learning-revolution/why-gpus-mattered/https://learn.physea.ai/history-and-context/early-ml-and-symbolic-ai/early-machine-learning/https://learn.physea.ai/history-and-context/early-ml-and-symbolic-ai/expert-systems/https://learn.physea.ai/history-and-context/early-ml-and-symbolic-ai/the-ai-winters/https://learn.physea.ai/history-and-context/early-ml-and-symbolic-ai/what-was-symbolic-ai/https://learn.physea.ai/history-and-context/key-milestones-timeline/deep-learning-breakthrough/https://learn.physea.ai/history-and-context/key-milestones-timeline/early-foundations/https://learn.physea.ai/history-and-context/key-milestones-timeline/the-llm-era/https://learn.physea.ai/history-and-context/open-questions-ahead/can-ai-reason/https://learn.physea.ai/history-and-context/open-questions-ahead/reliability-and-hallucination/https://learn.physea.ai/history-and-context/open-questions-ahead/the-alignment-problem/https://learn.physea.ai/history-and-context/open-questions-ahead/the-road-ahead/https://learn.physea.ai/history-and-context/the-llm-and-agent-era/chatgpt-and-the-chat-era/https://learn.physea.ai/history-and-context/the-llm-and-agent-era/gpt-3-and-few-shot-learning/https://learn.physea.ai/history-and-context/the-llm-and-agent-era/scaling-laws/https://learn.physea.ai/history-and-context/the-llm-and-agent-era/tool-using-agents/https://learn.physea.ai/history-and-context/the-transformer-moment/attention-is-all-you-need/https://learn.physea.ai/history-and-context/the-transformer-moment/bert-and-gpt/https://learn.physea.ai/history-and-context/the-transformer-moment/why-it-changed-everything/https://learn.physea.ai/language-model-architecture/https://learn.physea.ai/language-model-architecture/attention-explained/multi-head-attention/https://learn.physea.ai/language-model-architecture/attention-explained/query-key-value/https://learn.physea.ai/language-model-architecture/attention-explained/scaling-and-softmax/https://learn.physea.ai/language-model-architecture/attention-explained/where-it-came-from/https://learn.physea.ai/language-model-architecture/attention-explained/why-tokens-look-at-each-other/https://learn.physea.ai/language-model-architecture/context-windows/long-context-and-lost-in-the-middle/https://learn.physea.ai/language-model-architecture/context-windows/what-fills-the-window/https://learn.physea.ai/language-model-architecture/context-windows/what-is-a-context-window/https://learn.physea.ai/language-model-architecture/context-windows/why-limits-exist/https://learn.physea.ai/language-model-architecture/embeddings-and-vectors/from-words-to-sentences/https://learn.physea.ai/language-model-architecture/embeddings-and-vectors/meaning-as-geometry/https://learn.physea.ai/language-model-architecture/embeddings-and-vectors/measuring-relatedness/https://learn.physea.ai/language-model-architecture/embeddings-and-vectors/picking-a-model/https://learn.physea.ai/language-model-architecture/embeddings-and-vectors/semantic-search-and-rag/https://learn.physea.ai/language-model-architecture/how-text-is-generated/logits-to-probabilities/https://learn.physea.ai/language-model-architecture/how-text-is-generated/one-token-at-a-time/https://learn.physea.ai/language-model-architecture/how-text-is-generated/sampling-controls/https://learn.physea.ai/language-model-architecture/how-text-is-generated/when-it-stops/https://learn.physea.ai/language-model-architecture/parameters-and-layers/layers-and-depth/https://learn.physea.ai/language-model-architecture/parameters-and-layers/weights-vs-activations/https://learn.physea.ai/language-model-architecture/parameters-and-layers/what-7b-70b-mean/https://learn.physea.ai/language-model-architecture/parameters-and-layers/what-is-a-parameter/https://learn.physea.ai/language-model-architecture/the-transformer/self-attention/https://learn.physea.ai/language-model-architecture/the-transformer/the-block-repeated/https://learn.physea.ai/language-model-architecture/the-transformer/three-families/https://learn.physea.ai/language-model-architecture/the-transformer/what-came-before/https://learn.physea.ai/models-directory/https://learn.physea.ai/models/https://learn.physea.ai/models/choosing-a-model/benchmarks-mislead/https://learn.physea.ai/models/choosing-a-model/capability-cost-privacy/https://learn.physea.ai/models/choosing-a-model/start-from-the-task/https://learn.physea.ai/models/choosing-a-model/test-on-your-own-data/https://learn.physea.ai/models/context-and-pricing/batching/https://learn.physea.ai/models/context-and-pricing/how-billing-works/https://learn.physea.ai/models/context-and-pricing/prompt-caching/https://learn.physea.ai/models/context-and-pricing/the-context-window/https://learn.physea.ai/models/frontier-vs-local-models/privacy-and-control/https://learn.physea.ai/models/frontier-vs-local-models/the-capability-gap/https://learn.physea.ai/models/frontier-vs-local-models/the-cost-question/https://learn.physea.ai/models/frontier-vs-local-models/what-they-are/https://learn.physea.ai/models/licensing-and-open-weights/open-weights-vs-open-source/https://learn.physea.ai/models/licensing-and-open-weights/permissive-vs-restricted-licenses/https://learn.physea.ai/models/licensing-and-open-weights/what-you-can-legally-do/https://learn.physea.ai/models/model-architectures/dense-transformers/https://learn.physea.ai/models/model-architectures/diffusion-architectures/https://learn.physea.ai/models/model-architectures/encoder-decoder-and-hybrids/https://learn.physea.ai/models/model-architectures/mixture-of-experts/https://learn.physea.ai/models/model-architectures/state-space-and-mamba/https://learn.physea.ai/models/model-architectures/why-architecture-matters/https://learn.physea.ai/models/model-families/frontier-families/https://learn.physea.ai/models/model-families/open-weight-families/https://learn.physea.ai/models/model-families/what-is-a-model-family/https://learn.physea.ai/models/model-sizes-and-parameters/memory-and-running-a-model/https://learn.physea.ai/models/model-sizes-and-parameters/mixture-of-experts/https://learn.physea.ai/models/model-sizes-and-parameters/size-and-capability/https://learn.physea.ai/models/model-sizes-and-parameters/what-a-parameter-is/https://learn.physea.ai/models/types-of-models/embedding-models/https://learn.physea.ai/models/types-of-models/image-and-diffusion-models/https://learn.physea.ai/models/types-of-models/large-language-models/https://learn.physea.ai/models/types-of-models/multimodal-models/https://learn.physea.ai/models/types-of-models/speech-and-audio-models/https://learn.physea.ai/models/types-of-models/the-major-model-types/https://learn.physea.ai/news/https://learn.physea.ai/open-source/https://learn.physea.ai/papers/https://learn.physea.ai/prompting/https://learn.physea.ai/prompting/common-failure-modes/ambiguity-and-drift/https://learn.physea.ai/prompting/common-failure-modes/hallucination/https://learn.physea.ai/prompting/common-failure-modes/how-prompting-goes-wrong/https://learn.physea.ai/prompting/common-failure-modes/lost-in-the-middle/https://learn.physea.ai/prompting/in-context-learning-and-examples/how-examples-steer-behavior/https://learn.physea.ai/prompting/in-context-learning-and-examples/what-in-context-learning-is/https://learn.physea.ai/prompting/in-context-learning-and-examples/zero-shot-and-few-shot/https://learn.physea.ai/prompting/prompting-basics/common-beginner-mistakes/https://learn.physea.ai/prompting/prompting-basics/give-context-and-format/https://learn.physea.ai/prompting/prompting-basics/iterate-and-refine/https://learn.physea.ai/prompting/prompting-basics/what-makes-a-good-prompt/https://learn.physea.ai/prompting/reasoning-and-chain-of-thought/lets-think-step-by-step/https://learn.physea.ai/prompting/reasoning-and-chain-of-thought/reasoning-models/https://learn.physea.ai/prompting/reasoning-and-chain-of-thought/what-chain-of-thought-is/https://learn.physea.ai/prompting/reasoning-and-chain-of-thought/when-it-helps/https://learn.physea.ai/prompting/structured-output/constrained-decoding/https://learn.physea.ai/prompting/structured-output/schemas-and-strict-mode/https://learn.physea.ai/prompting/structured-output/why-structured-output/https://learn.physea.ai/prompting/system-prompts/instruction-hierarchy/https://learn.physea.ai/prompting/system-prompts/system-vs-user/https://learn.physea.ai/prompting/system-prompts/what-it-is/https://learn.physea.ai/running-ai-yourself/https://learn.physea.ai/running-ai-yourself/hardware-and-vram/memory-is-the-gate/https://learn.physea.ai/running-ai-yourself/hardware-and-vram/the-4-bit-rule/https://learn.physea.ai/running-ai-yourself/hardware-and-vram/what-fits-on-your-machine/https://learn.physea.ai/running-ai-yourself/local-inference-basics/how-to-start/https://learn.physea.ai/running-ai-yourself/local-inference-basics/the-app-layer/https://learn.physea.ai/running-ai-yourself/local-inference-basics/the-runtime-layer/https://learn.physea.ai/running-ai-yourself/local-inference-basics/two-layer-stack/https://learn.physea.ai/running-ai-yourself/local-vs-cloud-tradeoffs/cost-and-capability/https://learn.physea.ai/running-ai-yourself/local-vs-cloud-tradeoffs/privacy-and-offline/https://learn.physea.ai/running-ai-yourself/local-vs-cloud-tradeoffs/the-hybrid-pattern/https://learn.physea.ai/running-ai-yourself/local-vs-cloud-tradeoffs/what-is-the-difference/https://learn.physea.ai/running-ai-yourself/model-size-calculator/calculator/https://learn.physea.ai/running-ai-yourself/quantization/quant-levels/https://learn.physea.ai/running-ai-yourself/quantization/size-vs-quality/https://learn.physea.ai/running-ai-yourself/quantization/what-is-quantization/https://learn.physea.ai/running-ai-yourself/running-on-a-mac/mlx/https://learn.physea.ai/running-ai-yourself/running-on-a-mac/model-size-ceiling/https://learn.physea.ai/running-ai-yourself/running-on-a-mac/unified-memory/https://learn.physea.ai/running-ai-yourself/serving-and-runtimes/llama-cpp/https://learn.physea.ai/running-ai-yourself/serving-and-runtimes/ollama-and-lm-studio/https://learn.physea.ai/running-ai-yourself/serving-and-runtimes/vllm/https://learn.physea.ai/running-ai-yourself/serving-and-runtimes/what-a-runtime-is/https://learn.physea.ai/safety-and-security/https://learn.physea.ai/safety-and-security/alignment-basics/constitutional-ai/https://learn.physea.ai/safety-and-security/alignment-basics/outer-vs-inner/https://learn.physea.ai/safety-and-security/alignment-basics/rlhf/https://learn.physea.ai/safety-and-security/alignment-basics/what-is-alignment/https://learn.physea.ai/safety-and-security/data-privacy/do-they-train-on-your-prompts/https://learn.physea.ai/safety-and-security/data-privacy/local-inference-as-privacy/https://learn.physea.ai/safety-and-security/data-privacy/retention-and-zero-retention/https://learn.physea.ai/safety-and-security/data-privacy/what-leaves-your-machine/https://learn.physea.ai/safety-and-security/evaluating-trustworthiness/benchmarks-vs-real-tasks/https://learn.physea.ai/safety-and-security/evaluating-trustworthiness/grounding-and-citations/https://learn.physea.ai/safety-and-security/evaluating-trustworthiness/why-confident-is-not-correct/https://learn.physea.ai/safety-and-security/evaluating-trustworthiness/why-models-guess/https://learn.physea.ai/safety-and-security/jailbreaks/common-techniques/https://learn.physea.ai/safety-and-security/jailbreaks/mitigations/https://learn.physea.ai/safety-and-security/jailbreaks/what-is-a-jailbreak/https://learn.physea.ai/safety-and-security/jailbreaks/why-it-stays-hard/https://learn.physea.ai/safety-and-security/prompt-injection/how-to-contain-it/https://learn.physea.ai/safety-and-security/prompt-injection/one-channel-two-intents/https://learn.physea.ai/safety-and-security/prompt-injection/two-forms/https://learn.physea.ai/safety-and-security/prompt-injection/why-it-is-unsolved/https://learn.physea.ai/safety-and-security/safe-deployment-checklist/before-you-ship/https://learn.physea.ai/safety-and-security/safe-deployment-checklist/limit-what-it-can-do/https://learn.physea.ai/safety-and-security/safe-deployment-checklist/watch-and-undo/