Hayden™

Amateurs pay to avoid risk. Pros pay for the right to take it. When you find an asymmetric bet, put on as much risk as you possibly can.

@_chenglou some jedi level alpha

My dear front-end developers (and anyone who’s interested in the future of interfaces): I have crawled through depths of hell to bring you, for the foreseeable years, one of the more important foundational pieces of UI engineering (if not in implementation then certainly at least in concept): Fast, accurate and comprehensive userland text measurement algorithm in pure TypeScript, usable for laying out entire web pages without CSS, bypassing DOM measurements and reflow

FWIW, accelerate or die applies to your career, too I wouldn’t have said that a year ago. But I am watching people be 50-100-200% more productive than their peers by using AI I’m not trying to be a doomer. Anyone can do it. But you have to start shipping. Now

Stoked to have won!!! Wall Street Bananas was inspired by games I played growing up like Pokemon GBA and Stick RPG. So fun to design and build!!!

Absolutely the case. Always I hear boomers say: "I tested this and AI couldn't do it!!!" I ask: "What model did you use to test it?" It's ALWAYS ChatGPT Free or $20/month Claude. GenAI token I/O is the ultimate "get-what-you-pay-for" proposition. You need to pay for MORE TOKENS or BETTER TOKENS if you want the best results.

@rohit4verse keep it coming! good work

@HaydenH36 Great questions by the way. Will try them all especially when we talk about the question regarding tool calls because currently in the agentic space tool calling at scale is really hard to maintain. Thanks for this crazy share. Appreciate it.

@bucketshopcap Always has been 

@rohit4verse x.com/haydenh36/stat…

Amazing alpha. Impossible to appreciate enough. One little secret when building the harness: Ask the agent why the environment is failing to support its objectives. You will be shocked at the feedback loop this self-help has enables. I have a loop that asks the agent on failure: - Did you have any issues with your tools or associated functions? - Were there any bugs? - Did you have the right context for the job? - What do you wish you knew before? - What was confusing? - Do you believe you did the best you could, given what was available? If not, how could we improve your tools or context to greatly improve the quality of the result? Magic.

fanciest discussion ever

@FT fanciest deal ever

FT Exclusive: The discussions between the companies in recent months, which have now ended, were over a merger of Unilever’s food business and Kraft Heinz’s condiments division, according to people familiar with the talks. ft.trib.al/bnJLC7B

how it feels coding by hand when you hit claude rate limit

The book is The Visual Display of Quantitative Information by Edward Tufte.

This 140 year old chart has the most elegant payoff I've ever seen. It's a French train schedule from 1885. Every train between Paris and Lyon in a single day, on one sheet of paper. Each line = one train Slope = speed Crossings = two trains passing Flat segments = station stops 100 years later the chart got 10x better. They overlaid a high-speed train in red. One glance tells you it’s 3x faster. That’s how good this chart is. No Figma. No Claude. No D3.js. Just pure creative thinking on paper. No technology will ever replace the ability to think clearly about a problem.

@AYi_AInotes shockingly simple but makes all the difference x.com/HaydenH36/stat…

做Agent开发的朋友们，AI skill的底层逻辑要变天了， 今天这个思路至少帮你省80%的维护时间。 之前我们做AI技能都是写死SKILL.md，改一次得手动迭代， 现在有个新范式：让技能自己观察自己的失败，自己迭代进化。 刚刷到cognee-skills的思路，点透了现在AI技能最大的痛点：你写的SKILL.md是静态的，但环境天天变，模型更新了、代码API变了、用户需求变了，技能偷偷失效了你都不知道。 我之前做AI工作流就踩过这个坑：某个技能跑了两个月好好的，突然开始疯狂报错，查了半天才发现第三方API改了返回格式，之前的静态prompt完全适配不上。 什么叫活的技能系统？五个核心步骤👇 #AI #Agent开发 #prompt工程

@coreyganim @openclaw it can honestly be this simple and make all the difference x.com/HaydenH36/stat…

Your @openclaw works great... until it doesn't. Codebases change. Models update. And your "perfect" setup starts failing silently. The fix is a self-improvement loop: 1. Observe → Log every run (success, failure, errors) 2. Inspect → Spot the pattern when failures pile up 3. Amend → Fix based on evidence, not guessing 4. Evaluate → Roll back if it didn't help Simplest implementation: → Create a `.learnings/` folder → Log errors to `ERRORS. md` with context → Log corrections to `LEARNINGS. md` → Promote recurring fixes to `AGENTS. md` Now your agent gets smarter every time it fails.

@VadimStrizheus honestly it can be this simple and make all the difference x.com/HaydenH36/stat…

Read this article to make your agents self-improving. You’ll thank me later.

@aakashgupta it can even be this easy x.com/HaydenH36/stat…

I urge you to make your skills self-improving right now

@chadwahl it can be this easy x.com/HaydenH36/stat…

Love that people are just learning about the OODA loop with agents