Jonathan Chen

@ESYudkowsky .@ESYudkowsky why not make the book free online if it's so important?

"If Anyone Builds It, Everyone Dies" is now out. Read it today if you want to see with fresh eyes what's truly there, before others try to prime your brain to see something else instead!

@asmartbear @jasonfried NRR?

So I assume this is what @jasonfried was just arguing vehemently against. And I agree -- while Garry's way is a valid way, most people should go the Fried way. I was also explaining how there aren't only two ways. When unit economics are good, when retention is high (good both financially and indicates customers genuinely want the product), especially if NRR >100%, then continuing to invest in growth, even at roughly $0 profit or single-digit negative (if you want to take investment-- which doesn't have to be traditional either), then you are building wealth and derisking the business. If those assumptions aren't holding, then just "spend to grow" is high-risk. It's possible, of course, that your future self finds a way to fix everything. Just not a high-percentage bet.

@shl What's the reasoning behind this?

After 8 years of being fully remote, Gumroad is getting an office!

@lrhaughton It's possible to remove someone from following you.

People have disappeared. Twice now I’ve looked up a name. X says I don’t follow them. But I did. I know I did. I’ve sent and received DMs. I’ve commented on their posts. Yet X shows “follow” unselected. Not imposter version accounts. The real account. 🤦🏼

@damengchen Businesses are like babies. Never a better time than five years ago.

I probably chose the worst period to start my business - with a newborn at home. If I could redo it, I'd launch a business, exit before kids, enjoy 5 years with them, then start a new venture when they hit pre-K.

@dhh @Neovim No more Textmate?

I find that the beginner's eye only last for a moment. So I'll use this thread to record all the weirdness I encounter with @neovim and lazyvim out of the box.

@LBacaj Do you think IBM made some mistake?

Back in 1985, people were petrified by IBM's strength & dominance of the computer industry. Many believed as computers became networked, IBM was best positioned to soon dominate TV, Music, Telecommunication & other industries. It's fun to look back at how wrong they all were.

@BrianRoemmele Lee Kuan Yew would be proud.

Leadership. It looks like this. How Singapore prepares it citizens for the age AI. While some western countries sell yesterday’s careers to prop up institutions that refuse to grow and adapt with life long financial burdens.

@asmartbear Good ideas to reward them?

You never forget your first few customers. And the ones that stay with you for many years? They can’t know what their $49/mo meant to you. Everyone who bet on you early, in any form, holds a special place in your heart.

@catalinmpit Need resiliency as a backup though.

@dvassallo Have you considered moving south?

Living in the Pacific Northwest means that when the conditions are like this I must go outside to absorb the UV rays like a cold-blooded lizard.

@arvidkahl Is this like your own custom Mixtral mixture of experts?

Update on this. AI nerdery below, read at your own peril :P While I fell asleep last night, I had a revelation. I am in control of both the local AI AND the API that queries it. I can do what I want with the input AND the output, as well as the "transmission". So instead of looking for new models, I can just divide and conquer how I feed the existing LLM, even with limited context. Quickly built a text segmentation that splits my long texts into manageable chunks and then sequentially asks the same question of each chunk until either a positive response is found or I run out of chunks. With the LLaMa2 LLM I am using (currently, that's falkor-7b-q8_0.gguf), that's roughly 1.5k words per chunk, and that needs 3-5 chunks per 1h transcript. And my testing shows MASSIVELY improving results. Like VERY reliable Q&A on every single segment. Since I ask yes/no questions, I can stop at the first yes and ignore the rest of the segments. In my API, this was almost trivial to implement (although PHP SEVERELY lacks a nltk-like language processing library), so my transcription and inference backends now all run with much more effective inference, which I need for my "ask context-aware questions of podcast episodes to receive notifications on key terms and concepts." Thanks to everyone who replied to the original tweet, it loaded up my brain with alternative ways of thinking about this issue, and it came to me in a just-about-to-dream state of mind :D

@mityasmusin It’s the best of times. It’s the worst of times.

This year will be an extremely challenging time for many businesses. I don't envy anyone starting from zero right now.

@levelsio Seem better than TT to me. Maybe I haven’t invested enough time using TT though.

Deleted YouTube app from my iPhone Shorts are absolute cancer for my mind I already use the Chrome extension Unhook to block out all recommendations on laptop for years but on YouTube's iOS app you obviously can't block anything cause it's a native app not a site TikTok isn't even close as bad as this, you get lots of educational content, people talking about stuff, YouTube Shorts is just Instagram-type thirst traps but about 100x dumber I guess using YouTube in Safari iOS with 1Blocker I can make some content blockers to just hide Shorts there, will try later! inb4 "IT oNYL SHoWS You wHAT You wANT TO seE": get fukt m8

@LBacaj It’ll probably get better as token context increases and editors can edit multiple files. Seems simple compared to video generation!

Tiny observation about building software in the age of AI: I am still faster without AI when I know exactly what needs to be done for a feature. Especially when it touches multiple places in the code base. But I am a lot faster with AI when the feature is a little more vague.

@thesamparr Maintenance is easier?

I’ve been a serious weekend warrior weight lifter for past three years. In last 30 days I’ve switched to kettle bells and bands, no weights. My body feels great. Less achy. whereas I get banged up more with weights. Anyone made the switch. What happened?

@mityasmusin We’ve been talking deep fakes for a long time. We knew it would come to this.

I don’t need your bloody long reads on Sora to know we’re all doomed

@levelsio If you got a do over, would you work out more? Would you have similar results?

One counterintuitive thing about seeing all these entrepreneurs going to gym etc: I never went gym when I started doing startups, I didn't eat particularly clean But I did have lots of anxiety, depression then So I don't think it has any correlation with success in business, it's more like an after effect In fact I think the anxiety/depression can be a great fuel for your ambition, you trick yourself that when you're succesful and rich you'll be happy But of course you won't be when you reach success But that's when you finally do start gym and good nutrition and you actually get happy I think

@LBacaj Sensible agility

It turns out you can get very far in life, in careers, and in business just doing what makes sense each step of the way. Charlie Munger once said: “We never had a master plan and anyone that tried to do it we fired… We just did sensible things as we went along.”