有人问他:对于想在 AI 时代创业的年轻人,机会在哪里?他的回答很直接:做“应用层”,但不要只做包装器。Anthropic 每两三个月就发布一个新模型,每次发布都意味着一批之前做不到的事情变得可能。他说,Anthropic 的 API 业务一直在增长,而且不会被商品化——因为可能性的空间在不断扩大,总有新东西值得去做。
但他也说了一句很重要的话:护城河不是把 Claude 套个 UI,那不叫创业,那叫包装。什么意思?就是有些人做的事情,本质上只是:你问 Claude 一个问题,他帮你转达,然后把答案展示给用户看,中间加了个界面。这种东西,任何人都能做,三天就能复制,根本没有壁垒。
真正有价值的,是那些 Anthropic 做起来“不划算”的事情。比如:领域专业知识。大多数 Anthropic 的员工是 AI 研究员、产品经理,不是生物学家,也不是金融从业者。如果你在做生物技术 + AI,你对这个领域的理解本身就是壁垒;如果你在做金融行业的 AI 应用,那堆监管合规的知识,Anthropic 根本没兴趣也没能力去深耕。再比如:本地化的关系网络。这是他谈到印度 IT 服务公司时反复强调的一点——对本地市场的理解、客户关系、行业流程、交付经验,这些都不是一个通用大模型公司短期内能轻易复制的。
这次采访里,还有一个小细节我一直没忘。Dario 提到,他的一位联合创始人习惯把自己的想法和恐惧写成日记。有一次,他把这些内容喂给了 Claude,让 Claude 评论。Claude 不仅回应了日记里写到的内容,还说:“这是一些你可能有但还没写下来的恐惧。”然后——基本上都说对了。
I’ve never really had time to tinker with “Lobster” (OpenClaw). Today I forced myself to carve out an hour to install it and play around a bit. Honestly, it may not be as good as people online claim—some of the example use cases are things I’d already implemented two years ago, so in that sense it actually felt a bit disappointing.
At the end of the day, there’s nothing mysterious about it: as long as you have a machine/server that can run local models, even people who don’t know how to code can absolutely cobble something similar together—if they’re willing to spend time writing solid prompts.
That said, I do have to admit OpenClaw’s real value is that it makes “what’s possible” more imaginable and more reusable. It also gives people an extendable space—both a framework and a way of thinking. Add Skills into the mix, and it really does create that “wow, this is magical” first impression—especially with “memory.” In a way, it’s not that the product itself is magical; it’s that our personal imagination is often too limited. The recent explosive growth of Skills has suddenly made a bunch of scenarios feel practical—scenarios that previously felt almost unthinkable. In that regard, the product is genuinely impressive.
One last cold splash of reality: I wouldn’t recommend that the average person install it. Even setting aside hardware requirements, installation/debugging, and security concerns, network speed alone can be maddening. And if you really want to push OpenClaw to its upper limit, your wallet becomes a bottomless pit. For example, tools like Doubao or Yuanbao can do a decent job summarizing ordinary text—but without paying for more expensive models, the best you’re getting is basically a faster workflow, nothing more.
Eat fruit freely. What’s easy to overdo is juice, dried fruit, and smoothies.
Sugar isn’t “magic fat gain,” but it does make it easy to overshoot calories. If you want to lean out, reduce added/free sugars first.
Libido isn’t a health KPI. Low desire can reflect sleep, stress, meds, relationships, or hormones—but it’s not a simple “healthy vs. unhealthy” meter.
Many people eat because of boredom, thirst, stress, or cues. Before eating, ask: Am I hungry—or just triggered?
Exercise is one of the most evidence-backed mood tools. Constant negative news/doomscrolling can worsen anxiety and low mood—manage your “information diet” too.
Higher protein is often the easiest fat-loss strategy (more filling, more muscle-sparing), but “best” depends on your kidneys, preferences, and sustainability.
Eggs aren’t the enemy. For most people they’re a cheap, convenient, nutrient-dense protein source.
Breakfast isn’t mandatory. Eat it or skip it based on what helps you control hunger and total intake—not because someone said it’s “required.”
Getting your steps up is hugely underrated. Don’t worship “10,000”—but going from low steps to moderate/high steps often changes everything.
Genes influence your starting point, not your verdict. Lifestyle can shift risk dramatically—but don’t promise “almost everyone can completely override genetics.”
Use zero-calorie drinks as a buffer: water, sparkling water, tea, black coffee (if you tolerate it) can help you avoid impulsive calories.
You can’t spot-reduce fat. Training abs doesn’t mean you’ll lose belly fat specifically—your body decides where fat comes off.
If you want to get leaner, cut the least valuable calories first: sugary drinks, sweet coffee drinks, alcohol. And don’t “free-pour” cooking oil—measure it.
“Muscle confusion” isn’t the point. You need progressive overload + fatigue management, not random workouts for novelty.
Your self-labels shape your defaults. Replace “I’m lazy” with “I’m building a 3x/week training default.”
80/20 isn’t cheating. Eat like an adult 80% of the time, leave 20% for pizza/ice cream/burgers—as long as the big picture stays controlled.
Don’t demonize “oil.” What you should minimize is fried foods/restaurant hidden oils + total oil quantity. For daily cooking, it’s often smarter to use mostly unsaturated fats instead of piling on saturated fats.
If you can’t stay consistent, train when you have the most energy (often mornings). Done beats perfect.
3–4 days of lifting + lots of walking is usually more sustainable than grinding 6–7 days/week.
Don’t wait until you’re thirsty. In training and heat, avoid obvious dehydration (e.g., large body-weight drops) because performance suffers.
Enough sleep (usually ≥7 hours) is performance-enhancing. Fat loss, muscle gain, mood, and appetite control all start here.
Poor sleep makes you hungrier, less tolerant, and less likely to move. Many “fat-loss failures” are really sleep failures.
Get daylight early; keep evenings dim. Use your environment to train your circadian rhythm instead of relying on willpower.
Cold plunges/cold showers can be stress and mood tools, but they’re not a fat-loss hack. Icing immediately after lifting may blunt some hypertrophy adaptations—choose based on your goal.
“I’m too tired so I skipped” isn’t always laziness. Chronic fatigue? Check sleep, recovery, and training volume. Deload when needed.
Hit your protein first each meal, then decide whether you still want more carbs/dessert. It’s the simplest “anti-binge order.”
Protein target: for most trainees, about 1.4–2.0 g/kg/day (cutting can lean toward the high end). Don’t turn it into mysticism.
Carbs aren’t evil. The common fat-gain trap is the combo of high fat + high sugar + low fiber that drives overeating. Pair carbs with protein/fiber.
Food affects mood and focus—but mood and stress also affect food choices. Manage both directions.
Eat slower and stop around 7/10 full. Give your brain time to register “enough.” Don’t cling to fake “one hormone instantly does X” explanations.
After-dinner cravings: brush your teeth + leave the kitchen + use a zero-calorie hot drink beats white-knuckling.
More accurate version: fat loss = diet + daily activity, shape = lifting. Cardio is an accelerator, not a punishment.
Remove junk food from your home. If it isn’t there, you can’t mindlessly eat it.
Don’t train on vibes. Track your weights/reps/sets/technique. If you don’t track, progress is mostly guesswork.
Sleep isn’t fixed by “night mode” alone. What matters is brightness, duration, and stimulating content—long screen exposure at night can wreck sleep.
Daylight + being outside + moving often helps sleep more than any “sleep supplement” you can buy.
Obesity is an important risk signal, but “health” isn’t one sentence. Use waist size, blood pressure, glucose, lipids, and fitness to assess reality.
More muscle means a slightly higher metabolism—but the bigger differences usually come from how much you move and how much you eat.
Don’t wait for New Year or “next Monday.” Start today with something too small to fail (e.g., a 10‑minute walk).
You don’t lack motivation—you lack defaults: shoes by the door, workout clothes prepped, easy protein ready in the fridge.
Don’t treat plant-based meat like poison. Treat it like a processed food—check ingredients, sodium, saturated fat, and calories. Use less-processed proteins (beans/tofu) as your staple.