@Sentdex @jaminball Yes, eventually all these plans will meter by token
English
tower9000
1.7K posts
Apparently an unpopular opinion, but I don't think Anthropic owes anyone heavily subsidized tokens for their third party app.
Theo - t3.gg@theo
I can't help but feel personally burned by the Claude Code changes announced today. We put so much work into wrapping the (atrocious) Claude Agent SDK in T3 Code. It was the ONLY path they supported, so we made it work. It was hell. Now our users are getting their rate limits cut by 40x, despite us doing everything right. I listened to the Claude Code team. I had my issues with their direction, but I trusted them and took them at their word. I will never make that mistake again. Until we see significant change, it is safe to assume any statement from an Anthropic employee is a lie on a timer. The rug will be pulled, no matter how many promises are made beforehand.
English
English
bogosorting@bogosorting
qvr
English
@0xdoug @_notnotjake Workflows are changing pretty fast. Essentially all code reviews will go through an LLM first pass.
English
@_notnotjake Okay, give me a plausible alternative then? I’m not saying my assumptions are right, obviously something are wrong because Anthropic *is* at that ARR. But im trying to understand, in what way my mental model is wrong. So if you have a good explanation, it would be helpful to hear
English
I’m really struggling to see how the back of the envelope math on this works out…
There are generously 4 million characterized “software workers” in America. That’s pretty broad and includes a lot of people who aren’t really classical engineers don’t produce that much code.
That comes out to nearly $1k per month of average Claude spend across every dev in America. Yes, there’s some international usage, but it can’t be that much. Yes there is some non software Cowork usage, but that doesn’t use that many tokens. Yes, some non engineers are using Claude to vibe code, but I really doubt many are spending hundreds per month on.
Even if we assume 50% of all software workers are using Claude, that comes out to $2k spend per month per Claude user. Thats 10X more than the highest tier Max subscription. So almost all of Anthropics revenue has to be API billing
So the only explanation is that something like 20%+ of software engineers are not only Claude users but on API billing and regularly spending thousands per month. At $5/m Opus tokens that means the average API user has to be going through something like 25 million tokens per day.
*OR* the other possibility is API revenue is heavily power law dominated. Maybe there’s just something like 100k super users who are making up the majority of the revenue. For that to work the typical super user would have to be spending on the order of $50k/month and guzzling nearly 1 billion tokens per day.
Tannor Manson@Futurenvesting
Anthropic is now showing off $44 BILLION in annual recurring revenue. This is up $14 billion (+46.6%) since last month! BULLISH for AI Infrastructure $NVDA $AMD
English
This is incredible:
Elon Musk will receive 200 million super-voting shares in SpaceX ONLY IF the company establishes a permanent Mars colony with at least 1 million people.
In other words, Elon Musk will only receive this pay package if 1 million people live on Mars.
In other words, Elon Musk's biggest goal is now establishing a colony on Mars with a similar population as Dallas, Texas.
Musk is so optimistic about this goal that the vast majority of his pay is now contingent on it.
Life on Mars is closer than many expect.
English
When Tesla completes FSD, they need to do one thing: if you EVER paid for FSD on a Tesla, you get it for free on your next Tesla, since you've already paid for it and never got it.
As for me, I paid $8K for FSD in like 2014, "Coming Soon". I'm patient, but I have a long memory.
Nic Cruz Patane@niccruzpatane
Elon Musk on upgrading FSD hardware for customers who bought FSD on HW3 vehicles during today’s Q1 2026 Earnings Call: “Unfortunately, HW3, I wish it were otherwise, but HW3 simply does not have the capability to achieve Unsupervised FSD. We did think at one point it would have that, but relative to HW4 — it has only 1/8th the memory bandwidth of HW4, and memory bandwidth is one of the key elements needed for Unsupervised FSD, and it's just generally a thing that's needed for Al. If you're doing an order aggressive transformer, memory bandwidth is the choke point. For customers that have bought FSD, what we're offering is essentially a discounted trade-in for cars that have Al4 hardware, and we'll also be offering the ability to upgrade the car to replace the computer — you also need to replace the cameras, unfortunately, to go to HW4. To do this efficiently, we're going to have to set up micro-factories or small factories in major metropolitan areas in order to do it efficiently. I do think over time, it’s going to make sense for us to convert ALL HW3 cars to HW4 because that’s what enables them to enter the Robotaxi fleet and have Unsupervised FSD.”
English
tower9000 retweetledi
Our biggest resource of 2026 has been downloaded nearly 3000 times. It just got updated.
The 2026 playbook now includes a full breakdown of Meta's 31 March algorithm update, the rise of persona-led creative, partnership ads, AI UGC, and how the next wave of UGC is forming.
Four pillars: Meta, TikTok, AI, Creative. Built from £600M+ in revenue across the brands we run.
Retweet this post and comment "2026" and I'll send it across.
English
Cars sold in 2025: 5-10 million. Actual: 1.6 million (and declining).
EV revenue: $234-367 billion. Actual 2025 total revenue: $95 billion (and declining).
Bear case insurance revenue: $23 billion. Actual: Undisclosed, likely < $1 billion.
Insurance profit: 40% pre-tax margin. Actual: Undisclosed. 40% is impossible. Most definitely negative.
Human/autonomous ride-hail revenue: $42-327 billion. Actual: Zero
Gross margin: 43-50%. Actual: 18%
EBITDA margin: 30-31%. Actual: 12%
Enterprise Value/EBITDA: 14-18x: Actual: 117x (LOL)
Market Cap: $1.5-4.0 trillion. Actual: $1.3 trillion (312x P/E)
Free Cash Flow Yield: 4.2-5.0%. Actual: 0.5%
ARK then published increasingly more optimistic “research” reports on Tesla in the subsequent 3 years (before giving up). The final 2024 report projected a $7.0-10.9 trillion market cap and $300 billion of annual free cash earned by 2029. While Tesla’s’ stock did rise 59% in the subsequent five years from the 2021 report (to a 312x P/E), the ARKK ETF lost 43%. The fact that ARK still has more than $10 billion of fee-paying investor capital despite destroying the majority of the billions that cascaded in near its 2021 peak remains a mystery. Perhaps it’s the ongoing more than 100 appearances on CNBC. Or the live TV prediction of earning 50% returns per year. Cumulative management fees total more than $600 million. And counting.
English
ARK Invest published its 2021 “research” report on Tesla, thoughtfully estimating metrics on where the car company would be in five years. The results are in. Turns out Cathie & Co were just a tad bit optimistic in their forecasting skill. That was sarcasm. The scorecard: 1/
English
inference is very profitable and probably a good opportunity to understand some basic business math
1. companies buy long lived assets like GPUs. these are one time costs and the asset depreciates over time
2. once you own this asset, you can plug it in and produce tokens which you can sell. the cost of goods sold here can be very low and you might be making 90% margins at scale, this is why we say inference is profitable
3. then you also hire employees to do r&d work to improve your systems, come up with new models, expand the business
if you add these 3 up you end up with $0. you're not producing a profit because the business is growing and you're reinvesting it all buying assets or r&d to meet demand
if it's obvious to other people the business is working, you can raise money from them to accelerate all these numbers so they max out in 5 years instead of 25
so on paper you'll be "losing money" every year but that's because you want to make sure you lock down the opportunity before someone else
the bigger your market is the bigger this burn can be because it's a function of potential
so when you see these companies losing a lot of money it doesn't mean the whole concept of their business broken
it's possible they misjudge and overinvest on 1+3 and will suffer some consequences but fundamentally 2 does work
dax@thdxr
@d4m1n i'm a bit confused why so many people say api tokens are sold at a loss this isn't true - these models are incredibly expensive compared to the gpu time cost there's potential for 90% margin depending on the model
English
I will not send you guys DMs; I will make it public. My family has been planning this vacation for over 9 months. We paid for our suite on the cruise for a family of 5, totaling upwards of $7,000. We are supposed to arrive at the harbor by 2 p.m., with the ship sailing at 4.
We booked a super-early flight that cost a little over $1,400, and I paid $240 for extra luggage.
For this vacation, my wife and I cleared our calendars, with a running cost of at least $30,000 for the whole week.
Last night we tried to check in but realized the flight was overbooked, and even after checking in, we weren’t assigned seat numbers.
I also paid $80 for remote parking at DFW for 6 days, just for convenience.
We told my wife we had to get to the airport. Our flight was at 6:30 a.m., so we left home at 3:30 a.m. and arrived before 5 a.m.
The first attendant told us there weren’t enough flight attendants, so the flight was cancelled. We spoke to another one, who said it was a weather issue in Miami.
What I know is that the flight went to Miami, but because it was overbooked, they randomly cancelled some passengers.
What upsets me most is that my wife warned me about this airline. Also, my son is supposed to turn 9 on Wednesday, all of that is gone.
Royal Caribbean did not refund our money. My children’s spring break and my son’s birthday have been ruined just because we booked with Spirit Airlines.
Spirit Airlines@SpiritAirlines
@ekun_II We're sorry this has been your experience. It's not what we aim for! Send us a DM, and we'd be happy to help out.
English
Who is running the administration’s absolutely abhorrent social media strategy?
@jason@Jason
Seven Americans are dead, and over 100 are wounded. Thousands of Iranians have died, including over 100 schoolgirls. America might be responsible for those 100+ school girls. Whoever is making these, and whoever is approving them, just.... stop.
English
@jasonlk @buccocapital Jason. Atlassian can't even squash the 100s of bugs in their flagship product. AI will never be a thing for them
English
@buccocapital The real issue is they can’t even ship an industry-leading AI Agent and it’s March 2026
It’s worse than it looks at almost all public B2B companies
Cutting won’t help. Hiring won’t help. The core is rotten because it cannot adapt.
English
Unfortunately I think we'll see meaningful layoffs in software this year. And I want to explain why it's just air cover to call them "AI-driven layoffs", even though every company will do so.
Yes, AI makes companies more efficient. Developers and marketers can do more. CSMs can have a wider span of control. You can answer 70% of your tier 1 support cases with AI. But that's not really what's going on.
But two things are more elemental to the situation, and the actual driver:
1. Valuations have reset, with a totally valid and reasonable focus on free cash flow minus stock compensation. And the math simply doesn't math.
2. Many of these companies staffed up during COVID and never actually took their medicine and got fit. They thought demand would come back and it mostly hasn't. Not in the same way.
Illustrative example, to pick on two companies, Atlassian and HubSpot, that I actually really admire:
- Age: Atlassian is 24 years old. HubSpot is 19 years old - # of employees: Atlassian has 14k employees at $22B market cap. HubSpot has 9k employees at $15B market cap
- SBC-Adjusted FCF: Basically ZERO
That's right. After 20 years, the actual cash generated and available to shareholders is ZERO
I do think the owners of these businesses understand that is no longer tenable.
But they have two issues now:
1. The actual technical talent needs to get paid
2. Their stocks are down 60-70% from recent highs
So here's the situation: They need to start making actual money, they have to pay their tech talent, their dollar grants are going to have serious dilution consequences, and their cost structures are completely bloated for their current market cap, especially compared to more nimble competitors.
If they keep paying all of these people in stock, their dilution will continue and the stocks will continue to be punished. If they pay them all in cash, they will have no fcf.
TL;DR Layoffs are unfortunately the only true answer. They are coming. They will be credited to AI, and that will be air cover for the real problem.
English
@buccocapital maxtower.com/where-is-faceb…
They don't even need a model. They can accomplish their goals of diminishing competition by using one of the existing open weight models.
English
If only Meta could build a competent model, every software and pro services company would sign up tomorrow if Meta promised not to put out press releases destroying their business
Would choke off a critical part of the revenue (and distribution path) for the labs overnight
English
Atlassian interview (end to end). Real questions I got.
1. Recruiter screen (20 min)
“Why Atlassian?”
“What’s your notice period, location, comp?”
2. DS/Algo (60 min)
- “Given an array, find the shortest subarray with sum ≥ K.” (sliding window + prefix)
- “Design an LRU cache.” (HashMap + doubly linked list)
- Follow ups: time, space, edge cases, tests
3. Coding round 2 (60 min)
- “Merge intervals, return non overlapping result.”
- “Top K frequent elements.”
- Follow ups: complexity, alternative approaches, input constraints
4. System design (90 min)
“Design a Jira like issue tracking system.”
Covered:
- data model (Project, Issue, Comment, Status, Assignee)
- APIs (create/update/search)
- search (filters, pagination, indexes, Elastic)
- scale numbers (QPS, storage), hot partitions
- consistency (status transitions), idempotency
- async work (notifications, webhooks), retries, DLQ
- caching, rate limits, audit log
- observability (SLOs, traces)
5. Hiring manager (45 min)
“Biggest project you owned end to end?”
“Tell me about a prod incident you caused.”
“How do you handle disagreements in design reviews?”
“What would your ex manager say you need to improve?”
So if you’re prepping: don’t just grind LeetCode. Be sharp on tradeoffs, failure modes, and how you debug under pressure.
English
I do wonder how @cursor_ai feels about having made a partnership with openai, promoted and defaulted openai's models in Cursor, only for them to withold their coding model from them just a few versions later.
I was pretty dissapointed when I saw their CEO on stage with Sama pumping their models when that happened. Made a deal with a devil imo lol
English
@zachweinberg Claude code is mainly winning because tokens are heavily subsidized.
There's zero moat. The open weight Chinese models are already pretty good and maybe 6 months behind Anthropic. Open source versions of Claude code like open code are pretty good and maybe better.
English
Can someone ELI5 if Claude Code is going to destroy every other venture funded coding tool or not?
English
you realize how true this is when you're hiring
everyone is talking to the same 5 people
Jared Palmer@jaredpalmer
I have come to believe that there are actually only ~5000 truly elite software engineers in the world and they do the vast majority of the work
English