Jason Erickson retweeted
Karpathy 开炮了。
这位前 OpenAI 联合创始人、前特斯拉 AI 负责人,在一段 16 分钟的对谈里,把当下 Agent 热潮的底裤扒了个干净。
他说得很直接:Agent 的本质就是大规模蒸馏。你花大价钱跑的那个大模型,99.99% 的容量都浪费在了它压根用不上的数据上。听起来刺耳,但你仔细想想,事实就是这样。
然后他给出了一条真正能跑通的技术路线:先用大模型把能力蒸馏到小模型里,再给小模型配上正确的工具,最后用一个闭环把整个系统串起来。小模型、正确的工具、一个闭环,三件事,够了。
这里面最关键的一环,是闭环。
小模型本身并不可怕,可怕的是它一旦接入闭环,就能自我校正、自我改进,越跑越强。模型的参数量可以小,但只要反馈回路搭对了,它就能长出远超参数规模的实战能力。
所以别再为更大的模型交税了。真正让一个 Agent 变得危险的,从来都不是参数量,是闭环。
这 16 分钟的信息密度,比今年市面上任何一门 500 美元的 AI 课都高。看完之后,再去读那篇 Harness, Loops, Graphs,你就会明白:闭环就是那副马具,套上了,小模型才能真正跑起来。
Jason Erickson retweeted
One of my lead developers had a worthy #QuoteOfTheDay "I don't think I need AI to order a burrito" LOL
Jason Erickson retweeted
"Let’s plan the roadmap for next quarter” is the modern day equivalent of “let’s fax everyone the strategy.”
We're going into q4 planning and it makes my brain hurt. Roadmaps existed because software was expensive to build. You needed a bunch of people. Meetings. Someone moving little boxes around Jira for three months from now. It makes no sense and I'm trying to work out what we call the replacement. Maybe just "Continuous Compilation", ideas just compile to working product. I want a list of ideas that we want to continuously compile. That's the roadmap now.
The cheapest way to figure out whether something is a good idea is often to just build the fucking thing.
Here’s how I work now. I've done this weekly over the last few months and it works. My current napkin math is that I can usually build 6 months of old-school roadmap in a week. Sometimes more. I've also hit a wall sometimes. It happens.
When we have an idea, I usually 10x the ambition and then write a prompt, not more than a page long, that basically says:
- here’s the ambitious goal
- here are a few use cases
- here’s roughly how I think we should build it in terms of where it fits into code base, design considerations, etc..
- here’s how we’ll know if it actually works
- now go fucking surprise me
Agents build all night. Not even with a crazy factory or complex agent or agents. The base models are just too good for that now.
In the morning I look at it. Ok, they usually build in a few hours. Usually some of it is great. Some of it is stupid. Some of it exposes that my original idea was wrong.
And almost every time it gives me 5 new ideas I couldn’t have planned upfront. So we throw shit out. Who cares? We learned something.
Then we change direction and send them off again. It usually takes 3 rounds before the shape appears and we've learned about the gremlins and possibilities.
95% of the time I end up building something with 10x the scope of what somebody had carefully spread across the next quarter in Jira.
“What should we put on the roadmap for next quarter?”
My brain now translates that to: “Oh, you're stuck in the 1500s”
Jason Erickson retweeted
Dear @marvel @marvelstudios: Please, for the love of it all, give us a one-shot of Michael Pena as Luis doing a recap of the entire MCU saga leading up to @Avengers: Doomsday.
Not your fault. Still your responsibility.
That's the line I keep coming back to when people ask me about distraction, about weight, about any behavior they wish they could change.
I grew up clinically obese. That wasn't my fault. I was a kid. I didn't design the food environment, I didn't stock the pantry, I didn't invent the snacks engineered to override satiety.
But at some point, the story stopped being about what happened to me and started being about what I was going to do about it.
The same is true for attention.
Your phone was built by teams of very smart people whose job is to keep you scrolling. The apps use variable rewards, social approval, infinite feeds. That's not paranoia — that's the business model. It's not your fault the environment is hostile.
But it is your responsibility to do something about it.
This is where most of the conversation goes wrong. People pick a side. Either it's all personal willpower (blame the individual) or it's all Big Tech (blame the company). Both framings fail you.
Blaming yourself leads to shame, and shame leads back to the phone. Blaming the company leads to helplessness, and helplessness leads back to the phone.
The more useful frame is the one you'd use with a friend: I know this wasn't set up in your favor. And I know you're the only one who can change it.
Look for the discomfort you're trying to escape. Schedule what you'd do without it. Make the phone harder to reach than the thing that matters.
Not your fault. Still your responsibility. Both are true, and only one of them moves you forward.
Jason Erickson retweeted
I'm a cardiologist. I've spent twenty years inside American healthcare, and I'll say the quiet part first: the system is indefensible.
We spend roughly 18% of GDP on health care — the highest on earth — and have a lower life expectancy than countries spending half as much. Medical bills are a leading cause of bankruptcy. I have patients rationing insulin. I fight prior authorizations for medications that prevent heart attacks and imaging that finds cancer early.
Anyone defending the status quo isn't practicing medicine. They're profiting from it.
So why am I not for #MedicareForAll ?
Because I've watched what government-run health care actually does — from inside the exam room.
Denial doesn't disappear. It changes uniforms. Insurers deny by prior authorization. Single-payer systems deny by waiting list. Canada's median wait from referral to treatment has run over 25 weeks. The UK's NHS backlog has hovered around 7 million. When you eliminate the price signal, you don't eliminate rationing — you just move it from a claims department to a queue, and queues are invisible. Nobody sues a waiting list.
Medicare rates don't cover the cost of care. Hospitals lose money on Medicare and Medicaid and make it back on private insurance. Extend Medicare rates to everyone and that cross-subsidy vanishes overnight. Rural hospitals go first. Independent practices consolidate into hospital systems. The physicians I know would retire early, and the smartest students would go elsewhere. You can legislate coverage. You cannot legislate a doctor into existing.
The fraud problem is already staggering. HHS loses an estimated $100 billion a year. Investigators found a hotel in Los Angeles where every room was a registered hospice with zero patients — just addresses billing Medicaid. If we can't police the programs we have, scaling to the entire country isn't reform. It's a bigger target.
And the innovation math is real. Just this month: the first oral PCSK9 inhibitor approved, five Lp(a) drugs in late-stage trials, engineered immune cells that clear arterial inflammation. Americans overpay for drugs, and that overpayment subsidizes global R&D. Crush prices to European levels and something has to give. I'd rather fix the pricing abuse than gamble the pipeline.
What I'd actually do, and none of it requires a $30 trillion bet:
Real price transparency, enforced — you can't have a market where nobody sees prices.
Break the employer link so insurance is portable and you don't lose coverage with your job.
Expand HSAs and direct primary care — my colleagues doing it spend more time with patients and charge less.
Break the PBM middlemen inflating drug costs.
Site-neutral payments so a hospital can't charge triple for the same test.
AI-driven fraud auditing — flag every provider billing 10 standard deviations above peers and audit them.
Universal catastrophic coverage so nobody goes bankrupt from cancer.
Now the honest counterpoint, because you deserve it: the strongest case for single-payer is administrative. Private insurers run 15-25% overhead. Traditional Medicare runs 2-3%. Most published economic analyses do project net savings. And every peer nation with universal coverage spends dramatically less than we do. Those aren't talking points — they're real findings that my side has to answer.
I just don't believe the answer to a bureaucracy that denies your care is a bigger bureaucracy that can't be sued.
Fix the prices. Fix the fraud. Cover the catastrophes.
Don't hand the whole thing to the same people who couldn't notice 29 hospices in one hotel.
Jason Erickson retweeted
🌟 Make-A-Wish Hawaii’s WISH! Winter’s Glow silent auction is NOW OPEN!
Bid on incredible items and experiences while helping grant life-changing wishes for Hawaii children with critical illnesses. Every bid makes a difference — and all proceeds stay local. ✨
Preview & bid here:
👉 lnkd.in/gGVmCT-t
(Or visit wishgalahawaii.com)
Gala is May 16 at Sheraton Waikiki — don’t miss your chance to give hope! 💛
#MakeAWish #SilentAuction #WISHWintersGlow #HawaiiGives
Jason Erickson retweeted
Concern is Growing that the Line Stalls over Our Area, potentially resulting in a Prolonged Severe Threat
Jason Erickson retweeted
Opinion: Tennessee is not heeding its physical education guidance. Kids need more recess | Opinion theleafchronicle.com/story/o…
Jason Erickson retweeted
People say "Abolish the IRS, switch to a flat tax system"
I analyzed books, IRS data, and research papers to understand whether it's a good idea.
While flat tax may sound good on paper, it's actually a lot more complicated...
🧵
Jason Erickson retweeted
I have 3 kids, a full-time job, workout 4x per week, coach baseball, volunteer at church, and spend the majority of my weekend time with my family.
Here's what a normal week looks like 👀
The Simple Time Block Breakdown ⏰
🔴 Work = 40-45 hours, little early/late
🟢 Social = Church, kids baseball, football
🟢 Family = 90% of wknd time and all evenings
🔵 Workouts = 3-4x CrossFit, BJJ, plus playing with boys
A few things to point out and how I work around the inevitable changes of entrepreneurial work, raising 3 young boys, and staying connected to my wife and friends.
I stick to core hours at work between 10a-4p every day except Thursday (BJJ). My best work days are the ones where I go straight from a morning workout to the office. I stay dialed in and get a lot done, which carries to the rest of the week.
I avoid work on the weekends (hence the "event" note) but often write my newsletter late Friday or early Saturday when kids are sleeping. I'd rather work on the edges and keep my weekends free.
My wife and I set our workout times and stick to them. If I want to go to Jiu-Jitsu, then I need to go at 6am. If I'm tired, that's fine, but I don't get to go in the evening instead. I have my time and it's up to me to go.
Speaking of my wife - she's incredible. She homeschools the kids and runs the house day-to-day. There's tension, sure. We are always tired and often feel like we're pushing a boulder up a hill.
We would like more 1-on-1 and 1-on-none time. We're very focused on providing that balance for each other, but it's always a bit less than we planned for. But it's also a season of life with young kids that we are embracing for their joy and energy.
We try to bring friends to our activities, e.g. the farmer's market, workouts, hikes, the zoo, kids games, or host watch parties during football season. I call this "happy stacking" because I'm doing multiple things I enjoy at the same time. Like sitting in an old van at a brewery.
It's hard not to want to do MORE. I honestly want to work more than I do, there's so much energy and momentum in the business. But I also want to spend more time with my kids, and hell I want to train more than I do (CrossFit and BJJ).
That's the biggest challenge
- Fitting in everything I want to do
- Realizing I can't do everything I want to do
- Not comparing to other people's social posts
- Remembering I'm on my own path and timing
One more thing: I have been recognizing and rewiring old beliefs about what the perceived limitations of my own path are, i.e. "guess I'll make less money then."
It's striking a balance between gratitude for what I have (and there is so much to be grateful for) and openness to the opportunities that come my way.
Thanks for reading all the way down. I'd love to hear from other people (especially parents) on how they make a busy, ambitious, but still pretty balanced life happen. Rooting for you all!
Jason Erickson retweeted
1) The fact that Twitter is running well with headcount down significantly really matters.
Whether they admit or not, everyone in SV admires Elon.
A lot of venture funded CEOs are sending emails like this; inspired by Elon and taking drastic action.
Margins are going up.
Jason Erickson retweeted
Flux, Helm Top Cloud Native User Recommendations for Continuous Delivery Technologies https://nitter.gq/t.co/TsiWwDWZjF @CloudNativeFdn
Jason Erickson retweeted
The more you teach, the more you learn.
Whatever you know about business and investing, teach it others.
You'll get smarter, faster.
Jason Erickson retweeted
These Photos of Asteroid Ryugu from Successful MASCOT Landing Are Amazing goo.gl/VDVRaJ
Jason Erickson retweeted
Astronomers have pinpointed the orbit of an extremely distant Solar System object called TG387, also known as "The Goblin." And many are gleeful since this object could help in the search for the mysterious Planet X or Planet 9 theverge.com/2018/10/2/17922…
Jason Erickson retweeted
“Focus may just make you fall for Firefox again.” 😘
mzl.la/2wMUJjC































