https://fathomlayer.com The latest guides, glossaries, and news from Fathom Layer. en-us Sun, 04 Oct 2026 18:53:45 GMT https://fathomlayer.com/glossary/llm-inference-engine https://fathomlayer.com/glossary/llm-inference-engine Thu, 03 Sep 2026 19:52:15 GMT An LLM inference engine is the software that loads a language model's weights, runs the model to generate tokens, and serves the results to applications, usually over an HTTP API. It decides how GPU m... https://fathomlayer.com/glossary/2nm-process-node https://fathomlayer.com/glossary/2nm-process-node Thu, 03 Sep 2026 19:52:15 GMT A "2nm" node is a chip-manufacturing generation — not a physical measurement of anything on the chip. It is the industry's name for a density and efficiency tier. The IEEE's device-roadmap body (IRDS)... https://fathomlayer.com/guides/securing-a-self-hosted-llm-server https://fathomlayer.com/guides/securing-a-self-hosted-llm-server Thu, 03 Sep 2026 19:52:15 GMT Running a model locally with Ollama, llama.cpp or vLLM(/glossary/llm-inference-engine) is a few commands. Exposing that server safely is the part the quickstarts skip. The failure mode is consistent:... https://fathomlayer.com/glossary/secrets-management https://fathomlayer.com/glossary/secrets-management Thu, 03 Sep 2026 19:52:15 GMT Secrets management is the practice of creating, storing, distributing, rotating, revoking and auditing the credentials software needs to run — API keys, database passwords, tokens, SSH keys, certifica... https://fathomlayer.com/glossary/prompt-injection https://fathomlayer.com/glossary/prompt-injection Thu, 03 Sep 2026 19:52:15 GMT Prompt injection is an attack where text that reaches a language model is interpreted as an instruction instead of as content — because the model reads instructions and data through the same channel a... https://fathomlayer.com/guides/vllm-vs-sglang-vs-ollama https://fathomlayer.com/guides/vllm-vs-sglang-vs-ollama Thu, 03 Sep 2026 19:52:15 GMT The question stopped being "which model?" and became "which engine runs it best on my hardware?". Three names cover most of the decision. Here is how they differ in the way that matters — without benc... https://fathomlayer.com/glossary/ai-red-teaming https://fathomlayer.com/glossary/ai-red-teaming Thu, 03 Sep 2026 19:52:15 GMT AI red teaming is the practice of deliberately attacking your own AI system — with adversarial prompts, jailbreaks and edge-case inputs — to find where it breaks before someone else does. It borrows t... https://fathomlayer.com/glossary/fido2-webauthn-passkeys https://fathomlayer.com/glossary/fido2-webauthn-passkeys Thu, 03 Sep 2026 19:52:15 GMT FIDO2 is a set of standards for logging in with public-key cryptography instead of a password; WebAuthn is the browser API half of it; a passkey is the credential it creates. Together they are what "p... https://fathomlayer.com/glossary/mcp-tool-poisoning https://fathomlayer.com/glossary/mcp-tool-poisoning Thu, 03 Sep 2026 18:16:55 GMT MCP tool poisoning is an attack where a malicious or compromised MCP(/glossary/mcp) server hides instructions in the metadata a model reads — tool descriptions, parameter schemas or tool output — inst... https://fathomlayer.com/glossary/a2a-agent2agent-protocol https://fathomlayer.com/glossary/a2a-agent2agent-protocol Thu, 03 Sep 2026 18:16:55 GMT The Agent2Agent (A2A) Protocol is an open standard for letting independent AI agents — built by different teams, vendors or frameworks — discover each other and delegate work, without a bespoke integr... https://fathomlayer.com/glossary/deepseek-mla-and-deepseekmoe https://fathomlayer.com/glossary/deepseek-mla-and-deepseekmoe Thu, 03 Sep 2026 18:16:55 GMT Multi-head Latent Attention (MLA) and DeepSeekMoE are the two architecture choices that let DeepSeek-V3 run as a 671-billion-parameter model while activating only about 37 billion parameters per token... https://fathomlayer.com/glossary/ternary-llms-bitnet-b1-58 https://fathomlayer.com/glossary/ternary-llms-bitnet-b1-58 Thu, 03 Sep 2026 18:16:55 GMT A ternary LLM constrains every weight to one of three values — −1, 0 or +1 — which needs about 1.58 bits of storage each (log₂3 ≈ 1.58), against 16 bits for a standard model. The idea comes from Micro... https://fathomlayer.com/guides/local-llm-hardware-optimization-guide https://fathomlayer.com/guides/local-llm-hardware-optimization-guide Sat, 29 Aug 2026 18:47:13 GMT Local inference is memory-bandwidth bound, not compute bound. Generating one token requires streaming every active weight from memory exactly once. So the ceiling is: text max tokens/sec ≈ memoryband... https://fathomlayer.com/glossary/hnsw-algorithm https://fathomlayer.com/glossary/hnsw-algorithm Sat, 29 Aug 2026 18:34:00 GMT HNSW (Hierarchical Navigable Small World) is a graph index for approximate nearest-neighbour search: it finds vectors close to a query by walking a layered graph instead of comparing the query with ev... https://fathomlayer.com/glossary/quantization https://fathomlayer.com/glossary/quantization Sat, 29 Aug 2026 18:34:00 GMT Quantization stores a model's weights at lower numeric precision — 8, 4, or even 2 bits instead of 16 — trading a small, measurable quality loss for a large cut in memory and bandwidth. It is the tech... https://fathomlayer.com/glossary/kv-cache https://fathomlayer.com/glossary/kv-cache Sat, 29 Aug 2026 18:33:58 GMT The KV cache stores the key and value vectors the attention mechanism has already computed for every token in the context, so the model never recomputes them. It is what makes generation after the fir... https://fathomlayer.com/glossary/local-ai-appliance https://fathomlayer.com/glossary/local-ai-appliance Mon, 17 Aug 2026 19:40:01 GMT A local AI appliance is a self-contained computer bought mainly to run AI models on your own desk or network, rather than renting them from a cloud API. It is a descriptive label, not a standard indus... https://fathomlayer.com/glossary/microled-wearables https://fathomlayer.com/glossary/microled-wearables Mon, 17 Aug 2026 19:40:01 GMT MicroLED is a display technology in which every pixel, or every subpixel, is its own microscopic inorganic LED, so the panel makes its own light without a backlight. In wearables it matters in two dif... https://fathomlayer.com/guides/llama-3-vs-deepseek-local https://fathomlayer.com/guides/llama-3-vs-deepseek-local Mon, 17 Aug 2026 19:40:01 GMT Both are strong open-weight families you can run locally. The split is simple: DeepSeek's reasoning models think before answering (a visible chain-of-thought pass that costs tokens and latency), while... https://fathomlayer.com/guides/tesla-p40-local-ai-rig https://fathomlayer.com/guides/tesla-p40-local-ai-rig Mon, 17 Aug 2026 19:40:00 GMT The Tesla P40 is the cheapest route to a large VRAM pool: 24 GB per card on the used market for a fraction of a consumer 24 GB GPU. Two of them give you 48 GB for the price of one modern card. The cat...