37,832 results · Most relevant
Joe Rogan recommends finding something that really gets you excited and that you enjoy. He argues real motivation is finding a thing you love and just doing that thing, whatever it is, even playing chess. If you find a thing and run that thing down, keep doing it and dive into it, you can get pretty far in this life and have an enjoyable life — you just have to find out what your thing is.
Jack Wotherspoon recommends using managed agents instead of building agent infrastructure yourself: DIY setups require container/VM configuration, harness engineering, and networking/firewall/security work, which he describes as tedious, while managed agents reduce the whole process to a single API call.
To use a managed agent, you call the agent instead of a model: use the base antigravity agent and set environment equals remote. Jack emphasizes that environment=remote is the key thing telling the API you want a managed agent running in the cloud; after that you give it regular input as you're used to.
Jack Wotherspoon recommends running agent verification and file work in the cloud rather than locally. He describes that running verification on your own laptop can steal the browser and start running things for you, disrupting your work, and that cloud sandboxes avoid overwriting files or messing up your machine because everything runs as a process and returns the output. He also notes your laptop doesn't scale for agent workloads.
John Stecher recommends thinking through your eval set for testing agents so output is sensible, consistent, and repeatable, and says having good evals you can run consistently to measure system health is paramount to success. He explains why: as model versions change you get different answers and accuracy skew; if harnesses change you get different answers; if data changes you get different answers.
John Stecher says companies that want to succeed with AI agents must get their data story right: data locked in transactional systems rather than moved into something agents can access and manage will fail, so move data into an accessible, manageable, agent-recallable layer.
Dr. Søberg endorses not going extreme: you don't have to expose yourself very much to heat or cold to get healthy cardiovascular benefits, and you don't have to swim for half an hour in cold water — one to two minutes per session is enough. Overdoing cold or heat exposure exhausts the cells and increases the risk of cardiovascular disease.
Consequential creator advice requires independent verification before Today can promote it.
The speaker advises that your evaluation and your ambition must be beyond what the model's ability is today, so that you stress-test it rather than making it easy.
Jeff Smith recommends turning off arbitrary ad hoc SQL execution and limiting the agent to custom tools and reports. This is one way to lock down what agents can do once inside the database.
Jeff Smith advises that the best place to implement significant security for the database is at the database itself. If you build security only into your tools, a zealous engineer can pick up another tool and point it straight at your database, bypassing security. But if you put security at the database layer, more times than not, you're going to be okay.
Bakken recommends that of the two weekly threshold sessions, one should be short intervals (around 45 seconds to one minute) and one should be longer intervals (around six minutes).
Bakken recommends that runners test the golden zone (sub-threshold) in their training.
You need to own your own memory in an agent system; as developers and engineers of these systems, we need to actually own our own memory, and engineer the capturing of traces into a system that allows for continual learning to be possible.
Supriya Subudhi advises understanding why you are using AI in the first place — whether for long-term or short-term benefits — before adopting it, noting that many PMs receive company mandates to use AI without understanding why.
The speaker advises baking the entire journey of using AI into what people are already doing in the app rather than building a standalone feature people must find, and stresses that education and awareness are super critical for AI feature adoption. At Bumble this meant asking users for feedback when they were adding pictures, and prompting users who hit a wall finding matches to check the feedback in their profile and maybe change something.
hedged
Carson describes an additional QA process: a QA agent spins up a child of the builder to run QA; every builder records a video of the feature being used and passing tests (for UI-facing features you have to look at it); in Devin the built-in 'test' feature builds test cases, records a video, and gives green or red; a video that's all green means what you think is happening is happening and it visually looks good; if your agent harness lacks built-in video recording, you should build it — it's hard but worth it; you should spin up as many agents as you can as long as you're not holding up your merge queue, since it's cheap; a wonderful engineer named Dale recently joined and helped install the QA process.
Build a measure at every layer of the AI platform, so decisions can be made quickly and measured effectively.
Lakshya A. Agrawal advises that including domain-specific information in AI system prompts can make models perform much better, and GEPA can help discover that automatically.
Roman Ugarte advises building the best-in-class experience for things that are going to happen, translating AI capabilities into amazing products for customers as the models get stronger.
Erina Karati recommends approaching agent state problems systematically: run controlled experiments and keep only changes that survive measurement, rather than trusting one demo or endlessly hand-tuning prompts.
Choose sources → to get a library of your own and a daily brief.