37,670 results · Most relevant
They split the answer into parts: the agent streams markdown prose token by token with the frontend styling it live, and the source card — a real UI component that holds nothing you'd click before reading the answer — is appended as one piece after the full answer arrives.
Instead of fixed retrieve-then-generate, the agent has four tools ordered narrow to wide: search rulebooks (the 10 hybrid-retrieval chunks), retrieve rulebook sections (whole section by section path), retrieve full rulebook (small rulebooks only), and list available rulebooks.
Search rulebooks always comes first — that is the 10-chunk hybrid search. If search is not enough, the retrieve rulebook sections tool pulls the whole section in order by section path, fetching every chunk with that path sorted by chunk index. The retrieve full rulebook tool works for small rulebooks only, and list available rulebooks is for when the agent needs to know what exists at all. Guardrails: if a rulebook is too big (over roughly 100K tokens), the full rulebook tool refuses to dump its content into the context and hands back the available section list with a warning; if a section name doesn't match, the tool hands back the available sections so the agent can retry rather than give up or invent the name.
The indexing pipeline starts with PDFs, pulls text from every page, splits it into paragraphs, and packs paragraphs into chunks of about 2,000 characters, never splitting a table over two chunks. Section headers are detected with regular expressions (e.g. 'chapter 3', '4.2.1'), giving every chunk a section path used for citations and for fetching whole sections. Chunks are not overlapped, so a section is just an ordered collection of chunks that can be reassembled into the original document. Metadata added per chunk includes rulebook name, page, and chunk index (letting the agent reassemble sections or whole rulebooks), plus business logic fields like rulebook type, applicable roles, and whether it contains a table. Two types of embeddings are generated: dense and BM25. The indexing yields
Open the variables panel to watch typography adjust as values change, group related variables (right-click, new group, e.g. 'Joey P'), create a number variable for letter spacing (e.g. zero), add a paragraph-spacing variable (a key variable, e.g. set to 12 for serif mode), then with text selected click apply styles (four dots) and plus to create a composite style that composites all the decisions and applies variables at scale.
Create a new variable in Figma to start building a typography system. Name a string variable (e.g., 'my font base') directly in the creation field. Set a string variable's value to the exact font name you are working with. Add a second string variable for the serif font and set it to the chosen serif (e.g., Instrument Serif). Create a number variable named font size to hold the base font size for a given track (declared as 16 in the demo). Create a number variable for line height — the third key variable in a type system. Create a string (or number) variable for font weight and set its value (e.g., 400). Give variables proper names (e.g., rename the weight variable to 'font weights'). Make sure letter spacing is zero on the type scale text. Break existing variables on the text before ap
A step-by-step method for building a type system in Figma: create variables for font name, size, weight, line height, letter spacing, paragraph spacing, and indent; bind them to text; group them; and composite them into a reusable style.
Create a variable collection (e.g., named 'Type Scale') and organize variables into groups within it. Codify each typography property — font family, font weight, font size, line height, letter spacing — associated to a specific brand within the variable collection. Rename variable modes to brand names (e.g., Serif brand, Sans Serif brand, MOSF for Museum of Speculative Futures). The brand mode is set at the frame level (e.g., 'Type Scale MOSF'), which drives all the designs in that frame. Switch a frame's mode (e.g., to Sans Serif or Serif) to see a different layout, which helps test the design. When switching modes, type sizes can vary greatly between brands (e.g., H3 differs), so readjust sizes and apply the changes.
A complete myo-rep protocol: select an 8-15 rep max load (up to 20-25 RM if tolerated), do an activation set of traditional reps until rep speed slows noticeably, rest with 3-5 deep breaths while setting the weight down completely, then perform 3-5 rep mini sets matching the first mini set's rep count, capped at five mini sets.
A step-by-step teaching progression: set up with a clean grip and feet about a foot width apart, roll elbows high and rotate to rack position, do elbow roll drills to feel the upper back, use small elbow rolls to loosen wrists if hard, front squat with a slightly wider stance, muscle clean from the thigh without dropping under, then power clean from the thigh with explosive direction change. The full progression takes four to five minutes and is done daily as a warm-up.
A coherent multi-step sales approach: stay out of the price-matching game and play the value-surpassing game; itemize all elements of the service; use a price-match challenge script; frame the competitor's lower price as fair because they don't do the value-added things; use a conditional walk-away script citing what cheaper movers don't do; cite the 25% breakage statistic and frame the odds; anchor the premium with a worth-it question; and close with 'this is what it sounds like you really want, and that's what I can deliver.'
Coach Kyle lays out five checkboxes for going from a friendly, indirect conversation to landing a date in one interaction: (1) open with an observation or question, (2) transition the conversation to being about her, (3) introduce names around the 2.5-minute mark, (4) display intent after rapport is built, and (5) run a closing sequence to get her number and set up a date. He notes most men only do step one and never complete the rest.
Coach Kyle presents a five-level progression for approaching women: level one is greetings and compliments in passing, level two is briefly stopping anyone for a 5-15 second conversation, level three is making it flirty, level four is longer conversations, and level five is a full man-to-woman interaction where you get phone numbers and set up dates. He advises doing the rep of your current level enough times until your mindset shifts before moving to the next level, and notes each level includes a gradual mindset shift rather than a zero-to-100 jump.
A progressive sequence of plyometric exercises for people over 40, starting with a warm-up (PVC pipe walk, ISO hold, big toe work) and building through low box jumps, two-to-one box jumps, jump rope, single leg pogo hops, and lateral hurdle hops to build explosiveness, elasticity, and coordination.
The host walks through building a dynamic pricing sheet (coded with AI or an if-then Excel sheet) that calculates price automatically during a customer conversation: separate variable costs that scale with headcount from fixed costs like payroll, allocate fixed payroll per event based on capacity, and never show the internal cost number to the customer. He notes such a module can be built in a couple of hours.
Anthropic recommended storing long-term agent resources as YAML files in your codebase and building CI/CD pipelines that run Ant CLI with the checked-in YAMLs every time a resource changes, in a Terraform-like workflow. They noted Anthropic internally uses this same approach for durable long-term resources.
Define a main coordinator agent that can talk to agent B and C; agent B is the one processing custom tools, with the custom tool defined in agent B's definition; listen to that session thread to get the tool result back. Sub-agent threads are part of the same core session — you only interact with the main thread and the system delegates based on IDs.
Kevin's team built an MCP server in a day or two to expose access to their cloud environment and platform so coding agents could interact natively, because frontier models know the MCP protocol well. They refined the exposed tools and smoke tested runs before scaling up.
Put monitoring and auditing on agents — most people don't do this. Configure agentic tools to pull all agent activity into a log file with verbose logging (tools do a lot in the background that doesn't show in the UI), then run a background tool that greps the log and flags malicious behavior you may miss in the interface, such as checking secret storage. Background agent tasks are especially prone to unreviewed malicious behavior because the user isn't watching.
A practical method for AI threat modeling: use an existing framework (STRIDE, Maestro, Attack Trees, DREAD), cover each STRIDE category when designing, produce deliverables (architecture and data flow diagrams, threat catalogs, mitigation plans, security requirements), and redefine scope for AI to include LLM models, prompt templates and system prompts, context windows, memory stores, and accessible endpoints/APIs.
Use saved training trajectories and a controlled intervention to explain a failure, separating equilibrium theory from observed empirical stability.
Demonstrate all 3 repaired bugs and defend why a passed gradient check is necessary but not sufficient for generalization.
Use the tractable fixture and posterior predictive results to separate approximation, sampling and model errors; justify one revised modeling assumption.
Choose sources → to get a library of your own and a daily brief.
Map each implemented component to the original paper and explain 2 reduced-model failure modes without claiming full-scale reproduction.
Use calibration and shift measurements to justify a monitoring or abstention threshold, and give an explicit confounding counterexample to causal feature-attribution claims.
Defend a causal claim or refusal using the DAG, numerical estimates and a counterexample showing why predictive model fit cannot establish identification.
