IF you're not AImaxxing its over

[iphone]

---

TL;DR: Infinite money glitch is live right now if you're willing to work. Use AI to code apps for small businesses. You have ~2 years before the window closes. If you're under 25 and not paying attention, you're cooked.

---

The Setup
Right now there's an infinite money glitch if you're willing to grind for it: using AI like Claude to code apps.
Think about it. All these looksmaxxing apps and sites are printing money — selling product to insecure LTNs and worse. You think the guys shipping these are elite programmers? No way. Most of them are just sitting down and having Claude write the code for them.
For the next two years there's insane opportunity to ship stupid little apps and pull $10–20K/month. All you gotta do is try.
If you don't know what Claude Code is, or what shipping an app even means, or where AI is headed — you're cooked. Hard filter: if you're over 120 IQ and not making >$300K/year by your mid-20s in this environment, you're missing it.

---

Why Now Is Different
Three years ago, to build a site or app and charge people money for it, you needed at least one mega-autist on the team — or you had to go full autist mode yourself. That bar is gone. It has never been simpler. I can't believe I'm even telling you guys this.

---

B2C vs B2B — Pick B2B
Two kinds of apps you can build:

• Consumer apps — selling to individuals
• Business apps — selling to businesses

Build for businesses. If your app helps someone make more money or makes their workday easier, they'll happily pay you $50–$100/month for it. Consumers won't.
Dumb-simple examples:

• Scheduling tool for plumbers
• Lead-gen for local service businesses
• Anything that helps a small business get more customers, get those customers to pay more, or get them to stay longer

The framework is just: pick a job, find the parts that suck, build a tool that makes one of those parts simpler. Then figure out how to ship it with AI.

---

No Ideas? Get a Real Job
If you can't think of anything, go work a side job in the trades or in some actual business that makes money. Use your brain. Watch the day-to-day. Find the friction. Then build something with AI that removes it.
It has never been easier.

---

Why 2 Years
If you actually pay attention to this stuff — SemiAnalysis, Dwarkesh, the other autists — you know the hundreds of billions getting poured into AI are going to keep accelerating capability. In two years, the paid frontier models will be better than 99.9% of humans at almost everything.
Right now if you can rub two brain cells together you can print money. In a few years the economy might be fully cooked and wealth inequality is going to look insane next to today.

---

If You're 16 and Planning on College
It's over. By the time you graduate 5–6 years from now, robots are going to be doing most of the jobs in the world.
Don't believe me? Open Twitter, search videos of the robots coming out of China. Then go open Claude and try to ship something.

---

The Real Lesson: Watch the Derivative
What matters isn't where AI is now. What matters is how fast it's improving. First derivative. Simple math. A lot of the progress is a straight line on a log plot — exponential.
You have two years max to get your shit together. After that, get ready to be part of the permanent underclass on government handouts.

 

[nwed]

Simple:

Right now there's an infinite money glitch if you are willing to work for it. Using AI and coding and claude to make apps.
Think about it, all these looksmaxxing apps and websites are making bank. Selling products to insecure ltn and worse. You think the guys who wrote those apps are super capable programmers and coders? No way.

Most of them just sit down and make AI, like Claude, code these things for them. For the next two years, there's insane opportunity to make stupid little apps for people and make ten, twenty K per month. All you gotta do is try.

If you don't know what claude code is, or what making apps is, or where AI's headed, you're cooked. Literally if you have over 120 IQ and you're not making > 300k / year after you're 16 you're missing out.

Three years ago if you if you want to build some website or make an app and start charging people money for it you needed at least one mega autist on your team or you needed to go autist mode yourself. It's never been more simple. I can't believe I'm telling you guys about this.

13 posts and vip lifetime?

 

[kingofearth]

Simple:

Right now there's an infinite money glitch if you are willing to work for it. Using AI and coding and claude to make apps.
Think about it, all these looksmaxxing apps and websites are making bank. Selling products to insecure ltn and worse. You think the guys who wrote those apps are super capable programmers and coders? No way.

Most of them just sit down and make AI, like Claude, code these things for them. For the next two years, there's insane opportunity to make stupid little apps for people and make ten, twenty K per month. All you gotta do is try.

If you don't know what claude code is, or what making apps is, or where AI's headed, you're cooked. Literally if you have over 120 IQ and you're not making > 300k / year after you're 16 you're missing out.

Three years ago if you if you want to build some website or make an app and start charging people money for it you needed at least one mega autist on your team or you needed to go autist mode yourself. It's never been more simple. I can't believe I'm telling you guys about this.

So where's your results then? Also, I agree.

 

[iphone]

So where's your results then? Also, I agree.

bro you want me to dox myself?

 

[kingofearth]

bro you want me to dox myself?

One demonstration of a result is not going to dox you if you analyse what you're sending across.

 

[iphone]

One demonstration of a result is not going to dox you if you analyse what you're sending across.

I can easily afford to throw away lifetime vip on this forum and you can't. QED.

Anything I could "Send" to prove anything is fakable. I could send a screenshot of my bank account and the millions I have but anyone with two brain cells knows that can be faked. etc.

 

[chudpiller]

You can’t just make an app and print money. Making the app is the easy part, it’s way harder to get users especially if the app is something that requires a certain concentration of users in a given area (e.g a local service finder).

 

[try2beme]

13 posts and vip lifetime?

Same q

 

[sonic55555]

Simple:

Right now there's an infinite money glitch if you are willing to work for it. Using AI and coding and claude to make apps.
Think about it, all these looksmaxxing apps and websites are making bank. Selling products to insecure ltn and worse. You think the guys who wrote those apps are super capable programmers and coders? No way.

Most of them just sit down and make AI, like Claude, code these things for them. For the next two years, there's insane opportunity to make stupid little apps for people and make ten, twenty K per month. All you gotta do is try.

If you don't know what claude code is, or what making apps is, or where AI's headed, you're cooked. Literally if you have over 120 IQ and you're not making > 300k / year after you're 16 you're missing out.

Three years ago if you if you want to build some website or make an app and start charging people money for it you needed at least one mega autist on your team or you needed to go autist mode yourself. It's never been more simple. I can't believe I'm telling you guys about this.

Basically there's two kinds of apps you can make. One targeting individual users and one targeting businesses. It's way better to make apps for businesses' cause if you're selling stuff or making people who make money with the app their life easier, they're way more willing to pay you fifty hundred dollars a month for whatever little tool that you made for them that makes their life easier. It's just simple shit like helping plumbers schedule appointments, finding new leads for people to sell their local business services to. All you have to do is think about the job yourself and think about how you could make it easier if you had some app or some service that would make some part of it simpler. Or how to get more customers for your users or how to get those customers to pay for more or how to get those customers to buy more things or how to get those customers to stay longer.

If you don't have any ideas, get a side job in the trades or some business that makes money and use your brain. Think about how you can bring technology to them. optimize their day to day workflows and shit. And then f figure out how to build this with AI.

It's never been easier.

By the way, the reason I said next two years Is simple. If you actually understand anything about this and follow people like semi-analytics, Dwarkesh and other autists, You'll know that hundreds of billions of dollars that everybody in tech is spending on AI is going to continue accelerating AI progress. I think in two years, the paid models will be better at almost everything than ninety-nine point nine percent of people. Right now, if you have a few brain cells and you could rub them together, you could make money. Legit in a few years the economy might be fucked, the wealth inequality is gonna be even crazier.

If you're like sixteen right now and you're thinking about I'm gonna go to university And I'm gonna get my degree. It's over. By the time you graduate, five, six years from now We're legit gonna have robots doing most of the jobs in the world.

If you don't know what I'm talking about, open Twitter and just start looking at videos of robots coming out of China. try to make some app with Claude or the other models.

What you gotta look at is the rate of change, how fast these things are improving, not where they are now. It's understanding the first derivative. Simple math. A lot of this progress legit looks like a straight line on a log plot. We're talking exponential progress. So you have two years max to get your shit together. Otherwise get ready to be part of the permanent underclass relying on government handouts.

i walk around with earbuds that feeds into Grok on my phone, so everytime someone asks me something Grok feeds me the answer into my ear and i repeat it out loud to the person. i call it Grokmaxxing

 

[iphone]

You can’t just make an app and print money. Making the app is the easy part, it’s way harder to get users especially if the app is something that requires a certain concentration of users in a given area (e.g a local service finder).

This is good feedback. The problem For most apps is distribution and finding people to pay for you. I totally agree with what you're saying because getting users and getting users to pay is the hard part. However, the issue is everybody reading this literally doesn't even have the willpower or desire to do this. So if you're not saying bro you can do this, nobody's gonna try.

This is why my biggest suggestion would be making some apps and websites that help businesses get customers. Or make more money from their customers. Or solve some other very painful problem.

And you could literally use AI to get the contacts of all these businesses in your local area, use AI to create a marketing campaign, use AI to send them emails, use AI to generate landing pages for them, use AI to show them how they would actually use this product.

 

[chudpiller]

Since you seem to use AI a lot, do you use local lms or are you considering it?

 

[iphone]

Since you seem to use AI a lot, do you use local lms or are you considering it?

So the the the thing with local LLMs is you have to buy either Mac Ultra or something like that. And the payback period on that is like three four years. You need at least 24 but ideally like 96 or 128 gigs of RAM to run any reasonable local LLMs. So you're looking to drop upfront investment you know three, four grand on that. You can get literally a year of Claude Max 200X plan for 2.4 grand And then you're using like the state of the art opus models and if you're not retarded, it's probably enough to start making money off it.

If for some reason you have a ton of money to waste on local LM models, sure, but basically the drawback of using local models is that they're smaller, they have less tokens and parameters, and their tokens per second is a lot slower. And obviously they are overall weaker because they don't have as much chain of thought capability. They do less compute time thinking, and so you're just kind of wasting your time.

The easiest thing that everybody needs to be doing is biting the bullet, paying the twenty dollars a month plan to start with, Claude or whatever open AI codex, I'm not shilling either one of them, I just think Claude is better. and start understanding the power of what you're working with, with with the mindset that hey, I can make money with this.

 

[chudpiller]

So the the the thing with local LLMs is you have to buy either Mac Ultra or something like that. And the payback period on that is like three four years. You need at least 24 but ideally like 96 or 128 gigs of RAM to run any reasonable local LLMs. So you're looking to drop upfront investment you know three, four grand on that. You can get literally a year of Claude Max 200X plan for 2.4 grand And then you're using like the state of the art opus models and if you're not retarded, it's probably enough to start making money off it.

If for some reason you have a ton of money to waste on local LM models, sure, but basically the drawback of using local models is that they're smaller, they have less tokens and parameters, and their tokens per second is a lot slower. And obviously they are overall weaker because they don't have as much chain of thought capability. They do less compute time thinking, and so you're just kind of wasting your time.

The easiest thing that everybody needs to be doing is biting the bullet, paying the twenty dollars a month plan to start with, Claude or whatever open AI codex, I'm not shilling either one of them, I just think Claude is better. and start understanding the power of what you're working with, with with the mindset that hey, I can make money with this.

That’s kinda cope

I run a local lm on my $1500 pc that has 56gb of combined ram+vram

Imo they are plenty effective for whatever I need to do. (Which granted is not deep level coding)

Also you can install different models based on what you want to do (coding, images, etc) so having less parameters than a modern cloud based model is less of an issue.

But yea they are slower than the subscription ones obviously.

 

[Alienmaxxer]

i walk around with earbuds that feeds into Grok on my phone, so everytime someone asks me something Grok feeds me the answer into my ear and i repeat it out loud to the person. i call it Grokmaxxing

Do you actually do this

High IQ if so

 

[iphone]

That’s kinda cope

I run a local lm on my $1500 pc that has 56gb of combined ram+vram

Imo they are plenty effective for whatever I need to do. (Which granted is not deep level coding)

Also you can install different models based on what you want to do (coding, images, etc) so having less parameters than a modern cloud based model is less of an issue.

But yea they are slower than the subscription ones obviously.

The returns on intelligence are not linear. Sure, you're running a small quantized model and you're getting okay results. The difference in capabilities between what you can run in terms of your actual productivity versus what you can get paying two hundred dollars a month for claude max or whatever open AI subscription is worlds apart.

Of course it can write a simple solution for you, but the difference on intelligence is not linear. You understand that these AI labs will literally throw a few billion dollars into training these state of the art frontier models. The models, even though they're a mixture of experts, probably take a hundred, three hundred, a million dollars worth of GPUs to actually fit in memory. There's a reason why it costs so much money because it's worthwhile.

Again, the returns on intelligence are not linear. If we were gonna play chess and compete with each other and I was ELO 2000 and you were ELO 1800, Eventually I would basically take all your money.

The fact that you cannot multitask is a huge hit. The fact that you cannot get your results back really quickly is a huge hit. You have to think of making money as a constraint on your time.

It's amazing that you're messing around with local LLMs and don't stop doing it, but it doesn't mean that you shouldn't be using state of the art frontier models.

 

[nathan V]

yea exactly ai is basically your savior in 2026 jfl

 

[Noctiѕ]

---

TL;DR: Infinite money glitch is live right now if you're willing to work. Use AI to code apps for small businesses. You have ~2 years before the window closes. If you're under 25 and not paying attention, you're cooked.

---

The Setup
Right now there's an infinite money glitch if you're willing to grind for it: using AI like Claude to code apps.
Think about it. All these looksmaxxing apps and sites are printing money — selling product to insecure LTNs and worse. You think the guys shipping these are elite programmers? No way. Most of them are just sitting down and having Claude write the code for them.
For the next two years there's insane opportunity to ship stupid little apps and pull $10–20K/month. All you gotta do is try.
If you don't know what Claude Code is, or what shipping an app even means, or where AI is headed — you're cooked. Hard filter: if you're over 120 IQ and not making >$300K/year by your mid-20s in this environment, you're missing it.

---

Why Now Is Different
Three years ago, to build a site or app and charge people money for it, you needed at least one mega-autist on the team — or you had to go full autist mode yourself. That bar is gone. It has never been simpler. I can't believe I'm even telling you guys this.

---

B2C vs B2B — Pick B2B
Two kinds of apps you can build:

• Consumer apps — selling to individuals
• Business apps — selling to businesses

Build for businesses. If your app helps someone make more money or makes their workday easier, they'll happily pay you $50–$100/month for it. Consumers won't.
Dumb-simple examples:

• Scheduling tool for plumbers
• Lead-gen for local service businesses
• Anything that helps a small business get more customers, get those customers to pay more, or get them to stay longer

The framework is just: pick a job, find the parts that suck, build a tool that makes one of those parts simpler. Then figure out how to ship it with AI.

---

No Ideas? Get a Real Job
If you can't think of anything, go work a side job in the trades or in some actual business that makes money. Use your brain. Watch the day-to-day. Find the friction. Then build something with AI that removes it.
It has never been easier.

---

Why 2 Years
If you actually pay attention to this stuff — SemiAnalysis, Dwarkesh, the other autists — you know the hundreds of billions getting poured into AI are going to keep accelerating capability. In two years, the paid frontier models will be better than 99.9% of humans at almost everything.
Right now if you can rub two brain cells together you can print money. In a few years the economy might be fully cooked and wealth inequality is going to look insane next to today.

---

If You're 16 and Planning on College
It's over. By the time you graduate 5–6 years from now, robots are going to be doing most of the jobs in the world.
Don't believe me? Open Twitter, search videos of the robots coming out of China. Then go open Claude and try to ship something.

---

The Real Lesson: Watch the Derivative
What matters isn't where AI is now. What matters is how fast it's improving. First derivative. Simple math. A lot of the progress is a straight line on a log plot — exponential.
You have two years max to get your shit together. After that, get ready to be part of the permanent underclass on government handouts.

I tried asking Claude and loveable ai to write codes for websites for small businesses so I can sell it to them. All the local businesses I cold called all hung up on me or said not interested

 

[Noctiѕ]

What are some good stocks for me to invest in right now

 

[iphone]

Here's an example of how you can use AI. Literally asked Claude to give me ideas of B2B businesses websites you could build.
Do you know what question to ask, AI is already amazing. IF you sit and read cope threads about how you should fraud returns to target, it's over.


1. AI Voice Receptionist — Small businesses (plumbers, dentists, salons, law firms) miss calls constantly and every missed call is lost revenue. Build a voice AI that answers, qualifies the lead, and books appointments straight into their calendar. Twilio + ElevenLabs + Claude wires this up in a weekend. $200–400/month is an easy sell because the math is obvious: one missed job pays for a year of your tool.


2. Automated Review Funnel — After a job, SMS the customer. Happy ones get routed to Google reviews, unhappy ones get routed privately to the owner. Local service businesses live and die by their star rating. Charge $50–100/month and you'll never lose a customer.


3. Quote-from-Photo — Contractor snaps a pic of a leaky pipe, broken fence, dirty roof, busted AC. AI generates a branded PDF estimate in 30 seconds. Trades waste hours every week writing quotes longhand. Sell to single operators per-seat or per-quote.


4. Local Permit/Lead Scraper — Building permits, code violations, expired listings, and Zillow data are all public. Scrape, filter, and DM daily warm leads to roofers, pest control, contractors, solar installers. They'll pay forever because outbound is their bottleneck.


5. Invoice Chaser — Small businesses are owed money constantly and hate confrontation. Build a bot that escalates politely via SMS + email until the invoice is paid. Flat $99/month or take a cut of recovered cash. The ROI sells itself.


6. Local Social Media Autopilot — Restaurants, gyms, salons, barbers need daily IG/TikTok content and never post. Auto-generate captions, hashtags, and image overlays from their menu, schedule, or inventory. Zero work for them, $150/month for you.


7. Bookkeeping Pre-Sorter for Accounting Firms — Junior accountants waste 60% of their time categorizing QBO/Xero transactions. Build something that auto-sorts, flags anomalies, and drafts journal entries. Sell per-seat to firms — they have budget and they're already paying humans to do it.


8. Job Costing for Trades — Most contractors have no idea which jobs are actually profitable. They eyeball it. Build a tool that tracks materials, labor, hours, drive time per job and shows real margin. Half of them are losing money on specific customer types and don't know.


9. Cold Outbound Stack for B2B Agencies — Small agencies pay $1–5K/month for outbound sales tooling and SDRs. Build a tool that drafts, personalizes, sends, and tracks replies. Tie it to LinkedIn + email. Charge based on meetings booked, not seats — your incentives align and you can charge way more.


10. Insurance Claim Documentation for Contractors — Roofers, water damage restoration, auto body — all submit photo evidence to adjusters constantly. Build a photo → structured report pipeline that matches whatever the carrier wants. Saves hours per claim. Charge per claim and watch it scale.

---

Common pattern across all of these: you're not selling software, you're selling time back or money in. That's what makes B2B work. Pick one, ship it ugly in 2 weeks, get one paying customer, then iterate.

1. HOA / Small Property Manager Suite — HOAs and small landlords (under 50 units) are drowning in maintenance requests, violation letters, dues collection, and tenant comms. They use email and spreadsheets. Build a portal that handles all of it with AI for letter drafting and vendor coordination. Sticky as hell — once they're in, they never leave.


2. Veterinary Practice Backend — Vets are like dentists but with worse software and equally desperate clients. Appointment recall, vaccination reminders, AI-summarized visit notes, pet history. The vet space is consolidating fast (private equity rollups) and independents need ammo to stay competitive.


3. Small Church / Religious Org CRM — Sounds dorky but there are 350K churches in the US, most under 200 members, all running on Excel and goodwill. Member management, donation tracking, sermon/event coordination, automated outreach to lapsed members. Insanely low churn — these orgs don't switch software.


4. Owner-Operator Trucking Compliance — Solo truckers and small fleets need to track DOT compliance, IFTA fuel tax filings, HOS logs, maintenance records. Build a mobile-first tool with photo capture for fuel receipts and AI-generated quarterly filings. Truckers pay because non-compliance = fines and lost CDL.


5. Funeral Home Software — Death tech. 19K family-run funeral homes in the US, software stack from 1995. Help them manage arrangements, generate obituaries, coordinate with cemeteries, handle family comms. Recession-proof and zero competition because nobody finds it sexy.


6. E-commerce Listing Optimizer — Sellers on Amazon, Etsy, eBay, Shopify spend hours writing listings and never optimize them. Point at a SKU, AI rewrites title/description/bullets for SEO, generates A+ content, A/B tests variants. Charge per-listing or per-store, scales to thousands of customers.


7. AI Hiring Pre-Screen — Restaurants, warehouses, call centers, trades — anything with high turnover wastes hours on phone screens. AI does the first-round interview by voice, scores candidates on attitude/availability/experience, sends top 3 to the hiring manager. Charge per screen or per hire.


8. Catalog / Spec Sheet Generator for Manufacturers — Small manufacturers and B2B wholesalers spend days every quarter producing product line cards, spec sheets, dealer catalogs. Point at their SKU database, AI generates branded PDFs, web pages, and PIM-ready data. Industrial customers have real money and zero modern tools.


9. Voice-to-Structured Daily Reports — Construction supers, field service techs, traveling auditors — anyone who works in the field hates writing reports. They talk to their phone for 60 seconds describing the day, AI generates a structured branded daily report with photos auto-tagged to locations. Per-seat to GCs.


10. Tutoring Center Back-Office — Kumon-style tutoring centers are a fragmented $20B market run mostly by immigrant family operators. They need worksheet generation, progress tracking, parent comms (often bilingual), and billing. Build the operating system for after-school learning centers. Massive TAM, zero modern competition, owners pay cash.

---

Pattern check: every one of these is a vertical where the buyer is already paying somebody (or losing money) to do this job manually, and the software they currently use is from the Bush administration. Those are the easiest dollars on earth.

 

[iphone]

I tried asking Claude and loveable ai to write codes for websites for small businesses so I can sell it to them. All the local businesses I cold called all hung up on me or said not interested

Did you make those websites already? Literally if you know what you're doing, you can get websites hosted for free on Cloudflare. literally for free. They will host websites on Cloudflare for free. You can make a really nice website for them to have some examples. You could automate this and to start out, don't call businesses, go in person. Once you sell a few, selling the next one is easier.

You gotta knock on 100 doors before one opens up, but it becomes easier.

Bring your iPad, bring your laptop, show them, hey, I need this website for you. For your business. Make sure that the website helps them make money. It's not about looking nice, it's about getting customers, right? And just give them some projection about how this website is gonna get them more customers because it's gonna generate leads.

It's sick that you're doing this. I promise you there's some little Indian kid out there who's making bank with the same exact idea. Don't give up.

 

[alexbrown8384]

Did you make those websites already? Literally if you know what you're doing, you can get websites hosted for free on Cloudflare. literally for free. They will host websites on Cloudflare for free. You can make a really nice website for them to have some examples. You could automate this and to start out, don't call businesses, go in person. Once you sell a few, selling the next one is easier.

You gotta knock on 100 doors before one opens up, but it becomes easier.

Bring your iPad, bring your laptop, show them, hey, I need this website for you. For your business. Make sure that the website helps them make money. It's not about looking nice, it's about getting customers, right? And just give them some projection about how this website is gonna get them more customers because it's gonna generate leads.

It's sick that you're doing this. I promise you there's some little Indian kid out there who's making bank with the same exact idea. Don't give up.

I was that little Indian kid doing ts a year ago lmfaoo

 

[chudpiller]

The returns on intelligence are not linear. Sure, you're running a small quantized model and you're getting okay results. The difference in capabilities between what you can run in terms of your actual productivity versus what you can get paying two hundred dollars a month for claude max or whatever open AI subscription is worlds apart.

I kinda feel like the inverse is true; the returns on intelligence are logarithmically diminishing. Everyone knows that the ai industry is not hitting the crazy jumps in intelligence that they once were, the same applies to basically every technology ever. Obviously a model with 100b parameters is not gonna be 10x smarter than one with 10b.

Of course it can write a simple solution for you, but the difference on intelligence is not linear. You understand that these AI labs will literally throw a few billion dollars into training these state of the art frontier models. The models, even though they're a mixture of experts, probably take a hundred, three hundred, a million dollars worth of GPUs to actually fit in memory. There's a reason why it costs so much money because it's worthwhile.

The ai labs spend so much because they are hyper scaling with investor and government funds, not because they are actually tuning a profit.

The fact that you cannot multitask is a huge hit. The fact that you cannot get your results back really quickly is a huge hit. You have to think of making money as a constraint on your time.

Yea fair. For brokecels like myself, the maybe one minute I save on a complicated response is not worth the cost for the models however. Also the fucking restrictions are annoying and with an abliterated model you can do whatever you want (not sure if Claude pro max is abliterated, but I doubt it.

It's amazing that you're messing around with local LLMs and don't stop doing it, but it doesn't mean that you shouldn't be using state of the art frontier models.

Yea fair enough mate.

 

[iphone]

I kinda feel like the inverse is true; the returns on intelligence are logarithmically diminishing. Everyone knows that the ai industry is not hitting the crazy jumps in intelligence that they once were, the same applies to basically every technology ever. Obviously a model with 100b parameters is not gonna be 10x smarter than one with 10b.

The metric that you want to look at is length of task that can be completed at eighty percent success rate.

Measuring AI Ability to Complete Long Tasks

metr.org

Returns on parameters are probably logarithmic. The returns on intelligence are nonlinear. You're conflating two different variables.

Because the real world is a semi-zero sum competition until we get to like post-scarcity sci-fi shit, Small percentage differences in win rates make for huge percentage differences in long term outcomes.

Think about casinos. They have a 0.5% (blackjack) to 6~% (slots) edge. While we are sitting here edging here to being rich, casinos made billions of dollars.

 

[iphone]

The ai labs spend so much because they are hyper scaling with investor and government funds, not because they are actually tuning a profit.

bro, 30 billion in ~2 years. BTW whoever's reading this, if you don't understand what log scale means in this chart you're not gonna make it

this is cope. Amazon didn't turn a profit for a decade. Profit != successful business

The reasons the labs are all in on training is cuz they see the future and understand investing is the way to win. Every single model individually has been profitable, its just that they are training the NEXT one

 

[iphone]

I'm gonna go lift but when I come back I'll be dropping some more truth bombs on you brokies

 

[Mrinfinityx]

---

TL;DR: Infinite money glitch is live right now if you're willing to work. Use AI to code apps for small businesses. You have ~2 years before the window closes. If you're under 25 and not paying attention, you're cooked.

---

The Setup
Right now there's an infinite money glitch if you're willing to grind for it: using AI like Claude to code apps.
Think about it. All these looksmaxxing apps and sites are printing money — selling product to insecure LTNs and worse. You think the guys shipping these are elite programmers? No way. Most of them are just sitting down and having Claude write the code for them.
For the next two years there's insane opportunity to ship stupid little apps and pull $10–20K/month. All you gotta do is try.
If you don't know what Claude Code is, or what shipping an app even means, or where AI is headed — you're cooked. Hard filter: if you're over 120 IQ and not making >$300K/year by your mid-20s in this environment, you're missing it.

---

Why Now Is Different
Three years ago, to build a site or app and charge people money for it, you needed at least one mega-autist on the team — or you had to go full autist mode yourself. That bar is gone. It has never been simpler. I can't believe I'm even telling you guys this.

---

B2C vs B2B — Pick B2B
Two kinds of apps you can build:

• Consumer apps — selling to individuals
• Business apps — selling to businesses

Build for businesses. If your app helps someone make more money or makes their workday easier, they'll happily pay you $50–$100/month for it. Consumers won't.
Dumb-simple examples:

• Scheduling tool for plumbers
• Lead-gen for local service businesses
• Anything that helps a small business get more customers, get those customers to pay more, or get them to stay longer

The framework is just: pick a job, find the parts that suck, build a tool that makes one of those parts simpler. Then figure out how to ship it with AI.

---

No Ideas? Get a Real Job
If you can't think of anything, go work a side job in the trades or in some actual business that makes money. Use your brain. Watch the day-to-day. Find the friction. Then build something with AI that removes it.
It has never been easier.

---

Why 2 Years
If you actually pay attention to this stuff — SemiAnalysis, Dwarkesh, the other autists — you know the hundreds of billions getting poured into AI are going to keep accelerating capability. In two years, the paid frontier models will be better than 99.9% of humans at almost everything.
Right now if you can rub two brain cells together you can print money. In a few years the economy might be fully cooked and wealth inequality is going to look insane next to today.

---

If You're 16 and Planning on College
It's over. By the time you graduate 5–6 years from now, robots are going to be doing most of the jobs in the world.
Don't believe me? Open Twitter, search videos of the robots coming out of China. Then go open Claude and try to ship something.

---

The Real Lesson: Watch the Derivative
What matters isn't where AI is now. What matters is how fast it's improving. First derivative. Simple math. A lot of the progress is a straight line on a log plot — exponential.
You have two years max to get your shit together. After that, get ready to be part of the permanent underclass on government handouts.

Do u know that 99,999% of the apps in the play store barely give u a salary and are not downloaded cause of low reputation ? U need to be extremely lucky in timing

 

[xandylooksmaxs]

i have read all of this and you seem well versed in ai. nicely articulated

 

[Mrinfinityx]

i have read all of this and you seem well versed in ai. nicely articulated

Big tech companies are already working on every field with mogger IQ algorithm analysts, which is certainly not the forum population, u will be getting pennies cause they will outscale u if even a niche is getting relevant (like Nestle in the agri-food), u just can't keep u to the top if u are not an algorithmic freak team with unlimited ressources, it is all useless, manual labor or social credits is the future for 99,999% of the bell curve

 

[iphone]

Big tech companies are already working on every field with mogger IQ algorithm analysts, which is certainly not the forum population, u will be getting pennies cause they will outscale u if even a niche is getting relevant (like Nestle in the agri-food), u just can't keep u to the top if u are not an algorithmic freak team with unlimited ressources, it is all useless, manual labor or social credits is the future for 99,999% of the bell curve

Bro, I'm not teaching people how to make a billion dollars. I'm teaching people how to make a million dollars. There's a huge difference. Long term AI means everyone's cooked, but for the next two years, you can make apps that print money, especially in the B2B space.

If you think you gotta train an AI model or something like that, its over for you. What you gotta do is use the AI models the labs spend billions on yourself.

 

[iphone]

Do u know that 99,999% of the apps in the play store barely give u a salary and are not downloaded cause of low reputation ? U need to be extremely lucky in timing

True and I need to be explicit. When I say app I mean mobile app and web app, not only app store apps. And even in the app store there's approx 100k apps making a hundred grand a year or more look up reports on that. As long as you are not a complete derp you can make some money even with mobile apps, tho you gotta make web apps for b2b stuff

 

[NoxInfernum]

Use Codex not Claude. Codex 5.5 very high thinking is absolutely freaking monster.

 

[ZyzzReincarnate]

—

 

[iphone]

Use Codex not Claude. Codex 5.5 very high thinking is absolutely freaking monster.

For the most part you're right. Codex 5.5 is better _right now_. The problem is it tends to think like a normie too much. Like when you talk to Claude and a and ask him, Yo should I get plastic surgery to Lookmaxx, It's not gonna give you Oh you're beautiful the way you are bullshit. It might, but you can straighten it out with what's called "in context learning".

Chat GPT is fine-tuned and RLHF to the most insane normie mindset you could imagine. So it's gonna give you really normy answers most of the time. And if you're trying to make something unconventional, you have to guide it even more so and fight with it all the time.

The difference between the two models isn't that big because they're both cutting edge frontier models. The difference between them and an on-device model you can run is big. Use codex, use Claude. Not a big difference. Gemini kinda eats shit for getting things done.

 

[NoxInfernum]

For the most part you're right. Codex 5.5 is better _right now_. The problem is it tends to think like a normie too much. Like when you talk to Claude and a and ask him, Yo should I get plastic surgery to Lookmaxx, It's not gonna give you Oh you're beautiful the way you are bullshit. It might, but you can straighten it out with what's called "in context learning".

Chat GPT is fine-tuned and RLHF to the most insane normie mindset you could imagine. So it's gonna give you really normy answers most of the time. And if you're trying to make something unconventional, you have to guide it even more so and fight with it all the time.

The difference between the two models isn't that big because they're both cutting edge frontier models. The difference between them and an on-device model you can run is big. Use codex, use Claude. Not a big difference. Gemini kinda eats shit for getting things done.

Bruh are u stupid codex has nothing to do with Natural Language Processing, it doesnt answer questions or not a chatbot. It codes extremely well and can do anything you want to if you dont exceed legality level.

 

[ZyzzReincarnate]

Find content​

Find all threads by iphone

 

[iphone]

Bruh are u stupid codex has nothing to do with Natural Language Processing, it doesnt answer questions or not a chatbot. It codes extremely well and can do anything you want to if you dont exceed legality level.

Bro, are you retarded? Did I ever say natural language processing? literally inventing words I didn't say.

Do you know that Codex under the hood is just some scaffolding that pre-injects in context messages to make it code for you and is literally sending the same API calls to open AI servers as chatgpt and other LLMs? The only thing that makes it unique is that it can run arbitrary command line tools on your computer.

The underlying model that's running is fine tuned for coding, but it's still based on the pre-trained same family of LLMs as other models use. Also you are objectively wrong. It literally is a chatbot Just with some scaffolding and ability to run command line tools on either your machine or in a VM. You can talk to it and it will answer questions. Try it. Open codex and ask it to write a poem for you. This is literally the lowest IQ take I've ever seen. Do you think it's some magical different model that works in a different way? It's sending the same API requests to a slightly differently fine-tuned model, depending on which model you have Codex configure to use. It has some clever scaffolding to be able to fork, to run command lane tools to manage context, etcetera, but under the hood it's a chatbot, bro.

Literally you outed yourself as Low iq.

 

[AustrianMogger]

---

TL;DR: Infinite money glitch is live right now if you're willing to work. Use AI to code apps for small businesses. You have ~2 years before the window closes. If you're under 25 and not paying attention, you're cooked.

---

The Setup
Right now there's an infinite money glitch if you're willing to grind for it: using AI like Claude to code apps.
Think about it. All these looksmaxxing apps and sites are printing money — selling product to insecure LTNs and worse. You think the guys shipping these are elite programmers? No way. Most of them are just sitting down and having Claude write the code for them.
For the next two years there's insane opportunity to ship stupid little apps and pull $10–20K/month. All you gotta do is try.
If you don't know what Claude Code is, or what shipping an app even means, or where AI is headed — you're cooked. Hard filter: if you're over 120 IQ and not making >$300K/year by your mid-20s in this environment, you're missing it.

---

Why Now Is Different
Three years ago, to build a site or app and charge people money for it, you needed at least one mega-autist on the team — or you had to go full autist mode yourself. That bar is gone. It has never been simpler. I can't believe I'm even telling you guys this.

---

B2C vs B2B — Pick B2B
Two kinds of apps you can build:

• Consumer apps — selling to individuals
• Business apps — selling to businesses

Build for businesses. If your app helps someone make more money or makes their workday easier, they'll happily pay you $50–$100/month for it. Consumers won't.
Dumb-simple examples:

• Scheduling tool for plumbers
• Lead-gen for local service businesses
• Anything that helps a small business get more customers, get those customers to pay more, or get them to stay longer

The framework is just: pick a job, find the parts that suck, build a tool that makes one of those parts simpler. Then figure out how to ship it with AI.

---

No Ideas? Get a Real Job
If you can't think of anything, go work a side job in the trades or in some actual business that makes money. Use your brain. Watch the day-to-day. Find the friction. Then build something with AI that removes it.
It has never been easier.

---

Why 2 Years
If you actually pay attention to this stuff — SemiAnalysis, Dwarkesh, the other autists — you know the hundreds of billions getting poured into AI are going to keep accelerating capability. In two years, the paid frontier models will be better than 99.9% of humans at almost everything.
Right now if you can rub two brain cells together you can print money. In a few years the economy might be fully cooked and wealth inequality is going to look insane next to today.

---

If You're 16 and Planning on College
It's over. By the time you graduate 5–6 years from now, robots are going to be doing most of the jobs in the world.
Don't believe me? Open Twitter, search videos of the robots coming out of China. Then go open Claude and try to ship something.

---

The Real Lesson: Watch the Derivative
What matters isn't where AI is now. What matters is how fast it's improving. First derivative. Simple math. A lot of the progress is a straight line on a log plot — exponential.
You have two years max to get your shit together. After that, get ready to be part of the permanent underclass on government handouts.

Did read and it was shit

 

[iphone]

View attachment 5047365
Did read and it was shit

Thank you for your feedback, enjoy potentially being poor though!

 

[AustrianMogger]

Thank you for your feedback, enjoy potentially being poor though!

My man i surgerymax. After i am finished i dont care about my life.
In 2028 the moment i did my 2 LLs I am finished and free
I can be a beggar and then i would kms at 30 having reached everythign i wanted

 

[HltnPSL]

---

TL;DR: Infinite money glitch is live right now if you're willing to work. Use AI to code apps for small businesses. You have ~2 years before the window closes. If you're under 25 and not paying attention, you're cooked.

---

The Setup
Right now there's an infinite money glitch if you're willing to grind for it: using AI like Claude to code apps.
Think about it. All these looksmaxxing apps and sites are printing money — selling product to insecure LTNs and worse. You think the guys shipping these are elite programmers? No way. Most of them are just sitting down and having Claude write the code for them.
For the next two years there's insane opportunity to ship stupid little apps and pull $10–20K/month. All you gotta do is try.
If you don't know what Claude Code is, or what shipping an app even means, or where AI is headed — you're cooked. Hard filter: if you're over 120 IQ and not making >$300K/year by your mid-20s in this environment, you're missing it.

---

Why Now Is Different
Three years ago, to build a site or app and charge people money for it, you needed at least one mega-autist on the team — or you had to go full autist mode yourself. That bar is gone. It has never been simpler. I can't believe I'm even telling you guys this.

---

B2C vs B2B — Pick B2B
Two kinds of apps you can build:

• Consumer apps — selling to individuals
• Business apps — selling to businesses

Build for businesses. If your app helps someone make more money or makes their workday easier, they'll happily pay you $50–$100/month for it. Consumers won't.
Dumb-simple examples:

• Scheduling tool for plumbers
• Lead-gen for local service businesses
• Anything that helps a small business get more customers, get those customers to pay more, or get them to stay longer

The framework is just: pick a job, find the parts that suck, build a tool that makes one of those parts simpler. Then figure out how to ship it with AI.

---

No Ideas? Get a Real Job
If you can't think of anything, go work a side job in the trades or in some actual business that makes money. Use your brain. Watch the day-to-day. Find the friction. Then build something with AI that removes it.
It has never been easier.

---

Why 2 Years
If you actually pay attention to this stuff — SemiAnalysis, Dwarkesh, the other autists — you know the hundreds of billions getting poured into AI are going to keep accelerating capability. In two years, the paid frontier models will be better than 99.9% of humans at almost everything.
Right now if you can rub two brain cells together you can print money. In a few years the economy might be fully cooked and wealth inequality is going to look insane next to today.

---

If You're 16 and Planning on College
It's over. By the time you graduate 5–6 years from now, robots are going to be doing most of the jobs in the world.
Don't believe me? Open Twitter, search videos of the robots coming out of China. Then go open Claude and try to ship something.

---

The Real Lesson: Watch the Derivative
What matters isn't where AI is now. What matters is how fast it's improving. First derivative. Simple math. A lot of the progress is a straight line on a log plot — exponential.
You have two years max to get your shit together. After that, get ready to be part of the permanent underclass on government handouts.

mirin

 

[iphone]

My man i surgerymax. After i am finished i dont care about my life.
In 2028 the moment i did my 2 LLs I am finished and free
I can be a beggar and then i would kms at 30 having reached everythign i wanted

Good shit bro, but you wanna soft money max and soft lifemaxx after surgerymaxxing cuz you wanna have an easy life. Like trust me flying business class and having a sick car and living in super safe places and not having to think about how much shit costs or being able to buy anything you reasonably want is SICK

 

[AustrianMogger]

Good shit bro, but you wanna soft money max and soft lifemaxx after surgerymaxxing cuz you wanna have an easy life. Like trust me flying business class and having a sick car and living in super safe places and not having to think about how much shit costs or being able to buy anything you reasonably want is SICK

Also 2 years is kinda exagerated but most office jobs will get replaced in next 5-10 years.
Streaming is where its at. AI cant become a streamer as life content would be way to expensive to create.
Also my AMD investments are good

 

[iphone]

Streaming is where its at. AI cant become a streamer as life content would be way to expensive to create.

Oh I don't think this is true. Have you seen the latest Chinese models? I think in two years, bro, we're gonna have full time streamers doing IRL level shit. And Simps are gonna be donating like hell.

 

[AustrianMogger]

Oh I don't think this is true. Have you seen the latest Chinese models? I think in two years, bro, we're gonna have full time streamers doing IRL level shit. And Simps are gonna be donating like hell.

How should that even work? The computational load is too much for real time rendering and we arent seeing any drastic performance jumps in compute.
Best you can hope for is 2x in 5 years considering moorse law is long dead.

 

[iphone]

Also 2 years is kinda exagerated but most office jobs will get replaced in next 5-10 years.

I think ten years is way, way, way too far out. Again, look at that benchmark in terms of task success rate versus task length. They're gonna get way more reliable, bro. Way faster than even five years.

 

[iphone]

How should that even work? The computational load is too much for real time rendering

look at ByteDance's Seedance 2.0 specifically. Again, if it can render a 15 second clip in let's say five minutes, all you need to do is 20x the compute ot make it realtime. You already can basically do it with $$ and there's a lot of tricks to make it faster that will happen.

 

[AustrianMogger]

I think ten years is way, way, way too far out. Again, look at that benchmark in terms of task success rate versus task length. They're gonna get way more reliable, bro. Way faster than even five years.

At worst if i make it past 2028 and get my double LL i can just use my smv anyway.

 

[AustrianMogger]

look at ByteDance's Seedance 2.0 specifically. Again, if it can render a 15 second clip in let's say five minutes, all you need to do is 20x the compute ot make it realtime. You already can basically do it with $$ and there's a lot of tricks to make it faster that will happen.

20x the compute is at least 40 years away

 

[iphone]

20x the compute is at least 40 years away

bro you know you can like... have the gpus work in parallel right? that it already exists? the matrix computations themselves are already super parallelized across cores. this is a completely wrong belief that you need 20x compute on the same chip

you can literally scale that almost linearly ofc there's some nerd parallel scaling diminishing returns but for models like seedance its ez

ask chatgpt urself

 

[pinterest]

dndr ai generated post using ai takes away the learning aspect from you and is often long term worse than if you did it yourself