MethodThe team describes a layered approach to ambiguous characterization: use chemical intuition — whether a candidate phase is plausible given synthesis conditions and prior knowledge — to disambiguate phases with similar patterns; treat thermodynamics and physics as the biggest priors; when a phase matches nothing in papers or databases, use computational tools to infer what atomic configuration would explain the observed phases; run replicates to reduce ambiguity; and use multimodal characterization — XRD plus electrical, magnetic, and morphology measurements such as electron microscopy — to distinguish phases that look similar on XRD alone.
Method, not a verified protocol.
Latent.Space · Synthesis Superintelligence: from Semiconductors to Superconductors — Periodic Labs’ Liam Fedus and Ekin Dogus Cubuk
Anecdote · MoneyTara Sesha recounts that most enterprises were using chat platforms merely to ask questions while the agent revolution was happening, so OpenAI needed to get agents to enterprises as fast as possible — before they considered alternative products — and shipped a giant revolutionary change that broke all of their processes and conceptions about receiving updates, to unlock more value.
anecdotal
Lenny's Podcast · Stop planning for 2027: how OpenAI builds product 90 days at a time | Tara Sesha and Nan Yu (OpenAI)
Method · MoneyAt Factory, there is the concept of agent readiness, evaluated on eight pillars: (1) validation — do you have guardrails like linters and formatters in place? (2) build system — are your build commands in CI well documented? (3) feedback loops — are unit and integration tests tight? (4) docs — do you have readmes and agents.md files? (5) are dev environments cleanly reproducible such that an agent can use them? (6) is your code modular with clear boundaries and rules against sprawl? (7) observability — how long does it take to find out why something went wrong? (8) are you proactively security scanning so you don't have to worry about leaked secrets or vulnerabilities? If your agents and human engineers walk the same roads, paving them well is doubly useful.
Method, not a verified protocol.
AI Engineer · The Software Factory: From Bug Report to Production Code — Davis Palmie, Factory
MethodDuplicating an animated item and placing it inside another animated item creates nested layered motion; duplicating nested animated frames creates multiple layered animations; turn off clip content so nested items remain visible, and color-code nested items (red or blue) to distinguish them while building.
Method, not a verified protocol.
Figma · Figma for Edu: Prototyping with animations in Figma
RecommendationFor screen reader accessibility, you can use alt tags to explain animations.
Figma · Figma for Edu: Prototyping with animations in Figma
MethodThe host describes working with Figma MCPs configured with an LLM, copying a link to a selection (works in design mode too) to pass context, and prompting the LLM to take the animation from Figma, incorporate it into a web page, and give a breakdown of the animation.
Method, not a verified protocol.
Figma · Figma for Edu: Prototyping with animations in Figma
MethodA bouncing ball animation is created by having the ball go down and then copying and pasting it back, using ease-in and ease-out easing; loop the track to make the animation repeat continuously, and extend the animation track to slow down a fast animation.
Method, not a verified protocol.
Figma · Figma for Edu: Prototyping with animations in Figma
MethodMickey demonstrates copying a keyframe with Command C, moving the playhead to the target time, and pasting with Command V so the animation returns to its original position. He also notes that when copying keyframes to a new object, the pasted easing comes in linear and doesn't look good, so re-apply the ease from libraries.
Method, not a verified protocol.
Figma · Figma for Edu: Prototyping with animations in Figma
MethodFigma prototyping does not natively support scroll-driven animations, but you can press the Y key in design mode to add an annotation with additional context — for example, 'When scrolling to this section, perform the embedded animation' — and articulate section behavior such as six sections that animate when they come into view on scroll.
Method, not a verified protocol.
Figma · Figma for Edu: Prototyping with animations in Figma
Recommendation · MoneyMatt Owens advises not making shipping decisions an individual mandate — the team in charge and accountable for the work decides together when it's done. You know it's done when everyone says it's done and all the variables are looked at: time, money, sanity, creativity. Scope of work, project managers, rate cards, and time are healthy constraints that prevent endless refinement (like designing a million logos and never stopping).
Design Better · Matt Owens: Athletics co-founder on keeping your creative fire lit
Recommendation · MoneyWhen learning new technical skills with no formal resources available, Matt Owens recommends relying on research, books, and trial and error with a peer. He and his studio mate would 'just research as much as we could, learn books, whatever, trial and error.'
Design Better · Matt Owens: Athletics co-founder on keeping your creative fire lit
Recommendation · MoneyMatt Owens advises that one thing Athletics learned the hard way is that financial tracking matters: being able to forecast, understand where the money's going, where the work needs to go, and where the opportunities to grow accounts are.
Design Better · Matt Owens: Athletics co-founder on keeping your creative fire lit
Recommendation · MoneyMatt Owens recommends that when presented with an opportunity, you respond with 'how do we' rather than 'no'. He credits punk rock with giving him the fearlessness to start things — like putting out a record by a friend's band — without overthinking them.
Design Better · Matt Owens: Athletics co-founder on keeping your creative fire lit
Recommendation · SocialThe host recommends making concrete, impossible-to-ignore promises rather than vague commitments.
Charisma on Command · The 3 Traits That Make You Impossible to Ignore
Recommendation · MoneyFor new sessions without short-term memory, use long-term memory backed by a vector database. Store different types of memory in the vector database with different prompts that can understand your conversation. Use relational memory via entity graphs (e.g., Neo4j) with basic traversal to capture how things connect. Combine short-term memory, long-term vector memory, and graph memory in one agent architecture: short-term for the session, long-term for past memory, graph for relational data.
AI Engineer · Why Bigger Context Windows Won't Save Your Agent — Elizabeth Fuentes Leone, AWS
Recommendation · MoneyIsolate: separate context across agents. Share information between agents in a swarm via an invocation-state ID so memory pointers pass between agents without bloating context; when many agents need the same large data, share it via invocation state between steps.
AI Engineer · Why Bigger Context Windows Won't Save Your Agent — Elizabeth Fuentes Leone, AWS
Recommendation · MoneyCap tool invocations with a maximum tool count (e.g., 3 per tool) so the agent loop cannot invoke the same tool indefinitely, and require tools to return clear responses so the agent isn't left without a clear answer.
AI Engineer · Why Bigger Context Windows Won't Save Your Agent — Elizabeth Fuentes Leone, AWS
Recommendation · BodyWhen analyzing sprint contact, look at the overall position of the segments just before contact and during the first millisecond of contact; after initial contact, look at the overall contact phase and the maximum angles — maximum collapse and maximum trunk lean the athlete has to support.
Garage Strength · Why Most Athletes Train Speed Wrong w/ Sprint Scientist Dr. Jean-Benoit Morin
MethodA demonstrated method: use the provided instructions to identify what is taking up CPU in the sandbox, find the offending process ID, and kill it (e.g. 'kill 1250'); if you forget how, ask an AI assistant.
Method, not a verified protocol.
AI Engineer · How I Learned to Stop Worrying and Love the Sandbox — Matt Brockman, E2B
MethodFor the CPU runaway challenge: list processes to find what's taking up CPU, then kill the offending process by its process ID (e.g., kill 1250). If you forgot how to do something in Linux, ask your friendly AI or use the provided instructions.
Method, not a verified protocol.
AI Engineer · How I Learned to Stop Worrying and Love the Sandbox — Matt Brockman, E2B