Kalm

Building companies in public.

Documenting every product, workflow, and AI system I use.

Code • AI • Design • Content

[Zach Moskow @zachmoskow]

This will be a RUDE awakening. I’ve been saying this for months.

Claude Code, Codex, and every other coding subscription where the usage you get is INSANE will end and will end abruptly.

Providers are deliberately underpricing to capture usage.

That phase ends. And it ends soon.

These same $20–$200 plans will either jump sharply in price or introduce aggressive rate limits and overage charges. It will happen seemingly overnight as they prepare to go public.

The only people who will be fine are the ones who already built the habit of using at least 50% open-source models for coding. Whether is in opencode, Claude code with the @concentrateai api, or cursor.

[Robert Johns @RobertJohns]

One of the coolest uses of a text effect that I’ve seen in a long while. Bravo @hermeuscorp!

[Larry Fuqua @WizZifnab]

Now that Cursor is SpaceXAI, these two ladders should just be one.

SuperGrok includes Cursor Pro.
SuperGrok Plus includes Cursor Pro+.
SuperGrok Heavy includes Cursor Ultra.

Then you can stop calling Cursor out as a separate checkout. You buy Grok, you get the editor.

I'd bite on SuperGrok Plus tomorrow if it included Cursor Pro+. Right now I'm paying for the stack in pieces and the names don't even line up. @SpaceXAI
@cursor_ai

[Kalm @kalmedown]

@theo 4.6 is > than Kimi

[Suni @suni_code]

Why does Cursor feel more premium than Claude?

[All day Astronomy @forallcurious]

Humanity is for the first time witnessing plants breathing.

In other words plants are conscious beings!

[Astro Polo @astropol0]

SpaceX data might be the real cheat code for Grok 4.7....

most models are trained on internet text, papers, and code, while Grok 4.7 is a 2.1T parameter model getting massive supplemental training on actual SpaceX engineering data real systems, real failures, real constraints...

It will also come with better token efficiency than 4.6

thats the kind of data almost no other lab has access to

i think It could make it significantly better at real world engineering and systems thinking...

[Kalm @kalmedown]

@anupamrjp https://t.co/MD3OJ2mKyn & https://t.co/a9jTmcOOlm

[Linas Beliūnas @linasbeliunas]

All of Cursor's founders - 25-26 years old MIT dropouts - are now multi-billionaires following SpaceX's $60 billion acquisition 😳

Their story remains one of the wildest ones in tech ever:

→ 4 MIT students start a side project in 2022 (initially for CAD/mechanical engineering)
→ Pivot hard and create Cursor, the AI coding tool that developers love and & helped popularise vibe coding
→ Hit a $10B valuation
→ Decide copilots aren’t enough, so they move into agentic AI teammates & their own models
→ Need massive compute to compete with OpenAI & Anthropic
→ Meanwhile xAI is losing the coding race
→ Elon realizes catching up could take years, so he skips the line & buys Cursor for $60B
→ Cursor gets the world’s largest GPU fleet and compute, while SpaceXAI gets the coding leader
→ Founders become multi-billionaires; the first 50 hires ~$20-500M each

This is the beauty of startups at its finest.

[Kalm @kalmedown]

@designwkarthick You are a goddamn lie.

[Kalm @kalmedown]

@presleyacobb @sattyyouneed Cursor Cloud Agents would like a word about the “remote access is way way better” claim 👀

[Kalm @kalmedown]

@MaxBrinAI @sattyyouneed Stop it. Cursor > Codex

[Satyam @sattyyouneed]

Cursor is miles ahead of Codex.

It’s not even close.

[Kalm @kalmedown]

@g_est1 @JamesonCamp This is my current set up and I feel like I can code all day.

[Avengers Updates @AvengersUpdated]

Iron Studios has shared a first look at their life-sized Doctor Doom statue for ‘AVENGERS: DOOMSDAY’.

[The Batman Saga News @TheBatmanSagaNW]

Drone view of the set of ‘THE BATMAN: PART 2’ 🦇

(via: vyromedia_ | IG)

[The Batman Saga News @TheBatmanSagaNW]

The Batmobile sounds like a MONSTER on the set of ‘THE BATMAN 2’ 🦇❄️

📸 ross_sneddon | IG

[Kalm @kalmedown]

@Evinst3in @SirMacieel @SPAC89 Damn there goes my plan

[Kalm @kalmedown]

@trikcode Not with that water mark. Stop lying

[Kalm @kalmedown]

You said cache hit rates have been worse this week and that this is likely why some users are burning through usage much faster.

Can you please share the actual numbers?

• What was the average cache hit rate in the previous stable weeks?
• What is it right now for the affected users?
• How much does a drop in hit rate typically increase token consumption / quota burn?

Right now the explanation is still qualitative.

Since you previously said “Codex has nothing to hide,” the clearest way to prove that is to show the metrics instead of just describing them.

Looking forward to the data.

[NIK @ns123abc]

Thanks for the update Tibo, but the explanation doesn’t add up tbf

If cache hit rate is the issue, you should show us the metrics. What was the hit rate before? What is it now?

And how does that explain the users burning 70% usage?

If Codex has “nothing to hide,” show us the numbers. Updates help, but when the story changes every 24 hours, being transparent matters more than saying you’re transparent.