Feed every edit back into the AI project to refine voice
When you edit an AI draft, feed every edit back into the AI project so each subsequent draft sounds more like you.
40,353 results · Most relevant
When you edit an AI draft, feed every edit back into the AI project so each subsequent draft sounds more like you.
Notice when you're holding back — it might be suppressing a reaction, withholding your opinion, or avoiding a conversation altogether. Take a small, low-stakes step to stop restraining: ask the question, tell the joke, introduce yourself. When you make stopping restraining a habit, you will feel more and more comfortable expressing yourself more and more freely.
Set up your story in less than 30 seconds and only include details that will pay off later.
The narrator recommends describing your physiological state — racing heart, hyperventilating, bleeding, sweating — to convey intensity, listing specific physical tolls in sequence to convey exhaustion, and describing sensory details of your physical state (torn talismans, blood, sweat) to make the audience feel present in the scene; at high-tension moments, describe the protagonist's heart racing to convey fear and stakes.
The narrator recommends going into the characters you play — adopting their accent and gesticulating to increase tension — telling your story in the present tense, and sharing your internal monologue to create stakes that make the emotional ride feel more intense.
The narrator recommends introducing a specific detail early in your story that will pay off later (e.g., a skill or trait the protagonist has), and describing a character's physical details — build, appearance, demeanor — to build tension before a confrontation; when an antagonist appears, describe their physical presence before they act to build dread.
You don't need anything complicated for your hook. There is one hook that works in any situation: 'Oh, you're going to love this.' If you say it at the beginning of a story, the other person will immediately perk up. You need to follow it with the rest of the tips to actually deliver on your promise. An example hook is 'Truth that hit me lands like a butterfly and strikes like a lightning bolt at the same time.'
The source recommends that when your business requires you to be in rooms of power and influence but you're a maker — or requires networking when you're a craftsperson — you spend time doing something crafty in your private life so the artist in you gets a little airtime. More generally, find a place for the thing you love doing in the world, even if it's not in your business right now.
The source recommends hiring team members with different motivational profiles so you don't have to do all the things yourself — for example, if you're a craftsperson, hire a people person.
The creator offers a set of diagnostic questions: 'Are you someone who routinely sacrifices sleep?', 'Are you someone who is in a rocky place in your relationship?', 'Are you someone who's having trouble hiring?', 'Where's the mismatch? What's the problem?', 'What are the things that get in the way of your motivation?', and 'What are your strategies that have worked in the past that help you get back on track?'
A tool built with AI that pulls every video from a competitor's YouTube channel URL and runs each one through Ahrefs to estimate Google traffic, surfacing videos that perform well through Google SEO.
A tool built for an affiliate site operator that monitors pages, watches for ranking drops, and can be configured to send a Slack message when a post needs updating.
Anish Acharya describes a loop version of growth experimentation: every variant gets generated and measured; once results reach statistical significance with a high p-value, converge and ship that variant; then set up a long-term holdout and start the next experiment.
Anish Acharya sketches a bug fix loop: a bug report comes in, a repro is generated, a fix is created and reviewed; if high risk a human confirms before shipping to production, if low risk it ships automatically, and the customer can be emailed that the bug is fixed.
Anish Acharya cites Kavak, a used-car seller in Mexico, which runs a 'Jedi Academy' teaching everyone at the company, including mechanics, how to use new AI tools and technologies; at the end of the six-week course, employees ship a cutting-edge in-production agent.
Anish Acharya describes a pattern where an agent that gets stuck calls a human who coaches it through the problem; the agent then captures all the traces and learns, so it does not need to call again.
When you want to understand how a repository's primitives fit together, ask Codex to create a visualization of the repo in the style of an IKEA instruction manual. Charlie asked Codex exactly this for the Agents SDK repository.
For long-running tasks on a remote session, Gabriel Chua had Codex spawn a local thread on his machine and use the Chrome extension to inspect the app. The remote thread can send a message to the local thread instructing it to do click ops, browser work, and QA testing, then send results back; he notes remote-to-local thread coordination is worth considering for long-running tasks, and that a remote main thread can be instructed to test spawning a local thread and, if it succeeds, make that the default location for future local threads, especially for Chrome or end-to-end testing.
endorsed
Use custom instructions (agents.markdown in home directory) to set persistent behavior across all conversations.
Instead of describing each button press to Codex step by step (the slow way), show Codex the task by doing it yourself — record and replay. This applies especially to UI tasks.
Choose sources → to get a library of your own and a daily brief.