Zeiv

AI nerd
Exploring AI, agents & the future of tech.
Building in public.

[Noah @xbtnoah]

If you had watched this Jeff Bezos lecture in 2005, you might be competing with Amazon today.

Long before Amazon became the giant we know now, Bezos stood at Stanford and explained the principles behind its success. The entire 50-minute lecture is still available, and most people have never seen it.

He explains why Amazon deliberately made failure cheap, how an obvious website improvement actually hurt sales, and why he kept a feature that was costing the company money.

Then he recalls the moment Barnes & Noble, a much larger competitor, entered online retail. Bezos gathered his employees and told them they should wake up terrified. Not of Barnes & Noble. Of their customers.

He also shares a question that shaped Amazon's strategy: instead of trying to predict what will change in ten years, figure out what won't. Lower prices, more selection, and greater convenience.

The fascinating part is that Bezos wasn't looking back on a finished success story. He was explaining how he made decisions while Amazon's future was still uncertain.

Most people will scroll past this because it's 50 minutes long. Don't. Bookmark it now and watch it when you have an uninterrupted hour.

Follow me for more rare interviews and stories from the people who built the world's biggest companies.

[Zeiv @0xzeiv]

@xbtnoah wild that he was explaining all this in 2005. most founders only share their playbook once the outcome is obvious

[Zeiv @0xzeiv]

@xbtnoah The part about tearing down the office walls is underrated most companies say they want transparency then build an entire management structure around keeping people apart

[Zeiv @0xzeiv]

@xbtnoah The real question is whether Musk is a great investor or just the most successful survivor of leverage we've ever seen.

[Noah @xbtnoah]

Charlie Munger said this on camera, weeks before he died:

"I don't regard Elon Musk as truly rich, because I don't think it's sure that everything he's working on can work."

Then he went further.

"I would not invest in Elon Musk myself."

Munger's reasoning wasn't about Tesla's product or Musk's talent. It was about pattern.

"He's used leverage so much that he's doubled down right to the edge of extinction maybe two or three times."

Asked how many times Musk had walked that edge without falling in, Munger's answer:

"he's done it three times. Maybe he's got six more."

This is coming from a man who spent six decades building Berkshire on the opposite principle.

Munger and Buffett deliberately took smaller stakes than they could have afforded, using less leverage than was available, specifically so a bad stretch would never wipe out the people who trusted them.

Munger used the phrase "two hard pile" for things he'd rather not spend time thinking about, things he can't fix himself. He used it earlier in the same conversation for the risk of nuclear war. Elon Musk went in the same pile.

"I never met with him," he said. "As far as I'm concerned, he doesn't exist."

[Zeiv @0xzeiv]

@xbtnoah I’ve always liked Buffett, but Munger is a new discovery for me. Had no idea about him before this. Thanks for sharing.

[Noah @xbtnoah]

Warren Buffett and Charlie Munger explained an oil market problem in 2023. The 2026 crisis is showing why it matters.

At Berkshire Hathaway’s annual meeting, Buffett was explaining something easy to miss about American shale oil.

A new well can produce a huge amount at first. But its output falls quickly. Producers have to keep drilling just to replace the barrels that existing wells stop producing.

Then Buffett turned to Munger.

“It really dies fast, those shale wells. If you like quick death in your oil wells, we have them for you.”

Munger went on to describe how much technology it took to produce that oil at all. Companies had spent decades learning to drill deep underground and then miles sideways through rock.

They were not predicting a war or calling the price of oil. They were explaining why producing more oil is harder than watching its price go up.

That distinction matters in 2026.

Disruptions in the Middle East have restricted oil exports. Producers elsewhere have increased output, and alternative routes have helped move some barrels. But those responses have not fully replaced the missing supply. Inventories have been drawn down, while higher prices and shortages have also reduced consumption.

A higher price can make another shale well worth drilling. It cannot complete that well today, reverse the decline of existing wells, or instantly reopen a disrupted export route.

The market has three ways to absorb a shortage: find more supply, use stored oil, or consume less. When the first two cannot respond fast enough, more of the adjustment falls on the third.

Buffett and Munger were talking about the physical limits of oil production in 2023. Today’s crisis is a reminder that a price signal is not the same thing as an available barrel.

[Zeiv @0xzeiv]

@zakiraicoder There is something unique in Japaneese students they just to good

[Zeiv @0xzeiv]

i honestly don't get why people still let their AI forget them

the week GPT-6 Astra came out i gave it a folder and one rule: "write down everything you learn. reread it before every job"

the glass ball is that folder 19 days later. 165,000 leaves: 41,000 files, 118,000 calls, 6,000 notes to itself. one root, my first sentence

the idea is simple: you stop explaining yourself. it starts every job already knowing your lease, your bank and your printer, and gets faster every time

what Astra does with that folder:

→ reads it before it types a word. sorting my receipts: first time an hour, fourth time 6 minutes
→ asks a question and keeps working while it waits. ignore it, my mac beeps
→ runs a 2am routine on every note marked "later". a monitor port dead for a year got fixed on a saturday night, 4-minute flash
→ writes itself rules. after a subscription charged me twice: "check the bank export before anything renews". 14 mornings in a row since
→ hands boring reading to cheaper helpers, keeps the thinking. a full evening costs about 10% of my weekly limit

at first i thought the picture was a glitch. it's just 19 days of not forgetting

copy this today:

> open codex, pick GPT-6 Astra, effort on high. the $20 plan is enough
> make a folder called memory with a file called AGENTS.md. codex reads that file first, every time
> paste one line: "before every task read this folder. after every task write what you learned and what broke"
> drop in anything about your life: the lease, a bank export, the manual for the thing that never worked
> give it one job tonight and end with: "remember"
> tomorrow, ask it something you only ever said once

one folder and six lines. after that you never explain yourself twice

the catch: the folder gets fat and eats the weekly limit. every sunday it prunes dead branches, 20 minutes

the people doing this don't post about it. their Astra already knows their landlord and their bank. yours meets you as a stranger every morning

every chat you close tonight is a branch you'll never get back

[Argona @Argona0x]

https://t.co/K5FnOQbTdA

[Zeiv @0xzeiv]

@0xboan turns out the best trading decision the agent made was not trading at all

[Zeiv @0xzeiv]

@polydao wonder how much the fresh context review actually catches that a normal second pass misses

[Zeiv @0xzeiv]

@TeddyinMedia 100 accounts sounds useful until half of them are just repeating the same 3 posts

[Zeiv @0xzeiv]

@RohOnChain the real unlock is when agent starts killing strats that stopped working instead of just executing them

[Zeiv @0xzeiv]

@QuantIndicator @LuxAlgo 65$ is a bit pricey ngl

[Zeiv @0xzeiv]

@techwith_fahim Wallet selection is probably while strategy here copy trading is the easy part

[Zeiv @0xzeiv]

@adiix_official lets see how long it will stay free

[Zeiv @0xzeiv]

@defileo meanwhile 99% of people are still copy pasting into chat gpt

[Zeiv @0xzeiv]

@Shelpid_WI3M what if you put 2 agents on the same job though

[Zeiv @0xzeiv]

@kocer_eth sharp ! love it!

[Zeiv @0xzeiv]

@fluixoo depth is huge! well done