Jake Boyd

Opus 4.6 is the best model for vibecoding yet. I used it to build an entire workout coaching engine - but the story starts when Claude caught a training mistake two marathons' worth of wearable data couldn't. Let me explain... A year ago I started running seriously. Within twelve months I'd finished two marathons and was already planning my third. I was hooked - not just on racing, but on the whole system: periodization, heart rate zones, recovery optimization, the idea that if I could just get the inputs right, the performance would follow. So I did what every data-obsessed runner does. I bought a Whoop. I wore my Garmin on every run. I set up heart rate zones, built a training plan, and ran by the numbers. Easy runs in Zone 2. Tempo runs at threshold. I was disciplined about it. I thought I was doing everything right. I was training completely wrong for months. My lactate threshold heart rate was wrong. Not by a little - by a lot. Which meant every zone built on top of it was wrong too. My "Zone 2" wasn't actually Zone 2. My "tempo pace" wasn't actually tempo. I was running way too slow on almost every run, staying comfortable when I should have been pushing, and wondering why I wasn't getting faster despite putting in the miles. No app caught this. Not Garmin. Not Whoop. Not Strava. I had devices on both wrists and a chest strap and none of them said, "Hey - your zones are wrong and you're wasting your training." They just dutifully tracked whatever I did and told me I was being consistent. Great job. Gold star. You're consistently training at the wrong intensity. I only figured it out when I fed all my workout data into Claude and asked it to analyze my training. Within minutes it flagged that my heart rate zones didn't match my actual performance data. My threshold was significantly higher than what I'd been using. Overnight, all my zones shifted. Runs that used to feel like tempo efforts were actually easy pace. I'd been leaving so much on the table - and an AI caught it in minutes when months of wearable data couldn't. That's when it clicked: the problem isn't data. Runners have more data than ever. The problem is that most of us aren't coaches. We don't know how to interpret our own data, and when we get it wrong, nothing tells us. So I built Rex - so every runner can have access to that same intelligence, every single day. Rex is an AI coaching agent that looks at your actual workout data - pace, heart rate, effort, recovery - and makes real coaching decisions. Not a suggested workout. Not a chart you have to interpret. An actual call on what you should do today, with a clear explanation of why. Every morning you get a Daily Brief - your recovery state, today's workout, and the reasoning behind it. When your data says your zones are off, Rex catches it. When your HRV tanks, Rex adjusts the workout and tells you why. When you're ready to push harder, Rex pushes you. It's the coach I wished I had when I was running 40-mile weeks at the wrong intensity and thinking I was nailing it. I'm not a professional developer. I built the entire coaching engine with Claude Code - plan generation, readiness scoring, wearable data fusion, workout adaptation - all in production, mostly on nights and weekends while training for marathon number three. This isn't Garmin Coach (static templates that never adapt). This isn't ChatGPT ("make me a training plan" with no memory of who you are). This isn't a tracking app that shows you data and hopes you figure it out. Rex makes decisions. A human coach costs $200-400/month. Rex costs $25. I'm looking for 20-30 runners or hybrid athletes who use Whoop or Garmin to test it. If you've ever wondered why you're not getting faster despite doing everything "right" - Rex is for you. Sign up for early access at rexcoach.com, or DM me - happy to show you what a Daily Brief looks like.

@davieball DNAcomplete Essential - 1x WGS DNA Test Still waiting on file

@jakeboyd19 What did you order? Do you have the file yet?

Whole genome sequencing for $599, delivered in a format that makes AI answers 10–20× more accurate.

@patrickc @grok what’s the best way to get this done at a reasonable price?

I'm lucky enough to have a great doctor and access to excellent Bay Area medical care. I've taken lots of standard screening tests over the years and have tried lots of "health tech" devices and tools. With all this said, by far the most useful preventative medical advice that I've ever received has come from unleashing coding agents on my genome, having them investigate my specific mutations, and having them recommend specific follow-on tests and treatments. Population averages are population averages, but we ourselves are not averages. For example, it turns out that I probably have a 30x(!) higher-than-average predisposition to melanoma. Fortunately, there are both specific supplements that help counteract the particular mutations I have, and of course I can significantly dial up my screening frequency. So, this is very useful to know. I don't know exactly how much the analysis cost, but probably less than $100. Sequencing my genome cost a few hundred dollars. (One often sees papers and articles claiming that models aren't very good at medical reasoning. These analyses are usually based on employing several-year-old models, which is a kind of ludicrous malpractice. It is true that you still have to carefully monitor the agents' reasoning, and they do on occasion jump to conclusions or skip steps, requiring some nudging and re-steering. But, overall, they are almost literally infinitely better for this kind of work than what one can otherwise obtain today.) There are still lots of questions about how this will diffuse and get adopted, but it seems very clear that medical practice is about to improve enormously. Exciting times!

Seeing Japanese X on my timeline has been so beyond fascinating. It’s hilarious and I can’t stop reading it. But also it would be great to get Claude code and Clavicular back too

@justinskycak Skill issue

Pure vibe coding is a Ponzi scheme. Eventually, the technical debt comes due, and if you don't understand the foundations, you can't pay it off.

I know that codex writes better code than Opus but he’s so hard to talk to. It’s like speaking to someone with an accent I only half understand

Non-technical folks are uniquely positioned to unlock new ways to improve agentic coding. Not having the chops to solve a problem yourself requires you to figure out how to solve problems with agents. So fascinating

@Suhail The logical second order effect is that cost of code approaches $0, no?

We seem close to: - Give an agent access to a competitor app on a computer - Tell agent: Rebuild this app by using all its features - Agent tries app -> documents all flows/features/edge cases - The other agent builds all flows/features - They iterate trying/testing until done

This is crazy. Pretty sick move from Dario

@aakashgupta The reason math goes before writing is because math is logical and verifiable. Like code, it’s either right or it’s not. Writing isn’t. Models can write OK copy, but they definitely cannot write stories and they write mid Twitter and LinkedIn posts like this one.

Thiel has this completely backwards, and the data shows it. “Word people” aren’t surviving AI. They were the first casualties. Freelance copywriting agencies went from $600K in annual revenue to under $10K in 2025. Content writer job postings for digital marketing are projected to drop 50% by 2030. 81.6% of digital marketers already fear replacement, and that fear is justified because companies discovered “good enough” AI writing costs pennies compared to human salaries. The reason Thiel thinks math people have it worse is selection bias. When Microsoft lays off software engineers making $180K, Satya Nadella goes on stage at LlamaCon and tells Mark Zuckerberg that 30% of their code is AI-written now. Bloomberg files a report. TechCrunch writes three articles. 40% of Microsoft’s 2,000 Washington layoffs were software engineers, and everyone noticed. When a copywriter making $55K gets replaced by ChatGPT, nobody writes that Bloomberg story. One agency owner described losing all 8 employees including his sister. A gardening copywriter overheard her boss say “just put it in ChatGPT” six weeks before HR let her go. These stories ended up on personal blogs, not earnings calls. What Thiel is actually observing is a visibility gap. Coding displacement happens at companies with $3T market caps that file public disclosures. Writing displacement happens at 5-person agencies that just quietly stop invoicing clients. The math tells the real story: 276,000 tech workers lost jobs in 2024-2025. Content writer positions are declining 13% in AI-exposed fields. But the freelance writing market doesn’t have a Layoffs.fyi tracking it. There’s no WARN Act filing when a Fiverr client switches to Claude. Both sides are getting hit. The difference is that one side has a Bloomberg terminal tracking the damage, and the other side just has a Reddit thread.

The funny thing about this is if you’re still optimizing prompts in the traditional way you are not using AI right. It’s just engagement farming. Easy enough try on your own, output sounds good enough to feed the loop, not impactful enough to actually do anything

Finding users for my AI run coach: Reddit - -10 likes - top comment said “some people have no independent thought 😂” X - no traction Strava - 60 kudos and 7 waitlist signups from 1 post Meet users where they are I guess?

Want to know the stats behind your Strava data? This tool predicts your races scores and lets you know what’s working, what isn’t, and what the opportunities

@afrochicksnft @virtuals_io Sick! Im building this for running. Lmk if you want to try it 🤝

i'm building an ai fitness agent as a total beginner gymbot now works with openclaw! using @virtuals_io acp i was able to hire gymbot to coach me by paying it usdc

@pnwprincess23 Awesome post! Agree that most important thing is to just get miles and be consistent. That being said if people do need a training plan I built a website that makes customized ones based on their level / goals rexcoach.com

Every fitness app tells you what you did yesterday. None of them help you figure out what to do next. So I built one. Rex reads your Whoop/Strava/Garmin data every morning, makes a coaching decision, and tells you exactly what to run and why. Last week, it swapped my tempo run with an easy run for a different day because my HRV trend was declining. That’s the feature. The bad day is the product. Built in 35 days. Non-technical founder. 275 commits and 120k lines of code. Free training analysis (no account, just upload your data) 👇