Markizz

Building @my_kizzme | microSaaS for indie founders | No Legacy. All Signal.

[Markizz @mar_kizzme]

EKK 0.8 is now running in my setup.

I'm building a memory + workflow layer for AI agents: decisions, results and useful methods that survive the next chat.

The goal: make accumulated experience useful without turning every task into paperwork. 🧵

[Markizz @mar_kizzme]

A really cool competitor to Jev, Decider — can’t wait to test it out

[Base Labs @baselabs]

Today we release mammoth-1 and elephant-1!

We post-trained mammoth-1, the #1 ranked best decision model in the world, and elephant-1, which is the best model under 15B (and beats Jev).

We're open-sourcing all model weights and provide a technical report detailing all our training and data methods.

[Markizz @mar_kizzme]

DotS? Finally?

[Tibo @thsottiaux]

Dot, Dotter, Dottest

[Markizz @mar_kizzme]

Feels like they’re stepping on the same rake as Apple — but without the history or foundation to pull it off.

[Markizz @mar_kizzme]

OpenAI feels like it’s missing a product team that ships fast.

Codex turned out great. Why so little of everything else?

Not the core product work — just the flashy stuff like Grok Bots or Claude Motion early on.

They take products way too seriously.

[Markizz @mar_kizzme]

Dots bye?

[Claude @claudeai]

We’re opening a research preview of Claude Duet, a new app where Claude works as a proactive agent.

It comes to you with what needs your attention, completes tasks you ask of it, keeps working when you step away, and checks with you before it acts on your behalf.

[Markizz @mar_kizzme]

Already buying my 3rd Claude account.
The subagent waste is insane. That’s the price of high autonomy.
Also feels like the limits got cut hard.

[Markizz @mar_kizzme]

Honestly didn’t expect Grok Bots to pull ahead this hard.
I think Dots will catch up, but it’ll be a slow grind like Claude Code vs Codex.
Congrats, Grok earned it.

[Markizz @mar_kizzme]

Today we’re trading time for limits, and money has never felt this equal to time — more accounts means less steering.

[Markizz @mar_kizzme]

The math checks out on max reasoning.
In practice Opus still gets me further with less steering — it just understands the intent better.

Though with OpenAI’s resets right now, it’s still the better deal overall.

[Markizz @mar_kizzme]

Interesting how many users they lost in the 28 days after cutting limits.

The frequent resets aren’t random.

Still not ruling out them just adding more compute through other channels.

[Markizz @mar_kizzme]

Pure win for users.
More capacity, less bottleneck. Limits might even improve.

[Adam.GPT @TheRealAdamG]

https://t.co/4IoleXvt5J

OpenAI 🤝 GCP

[Markizz @mar_kizzme]

The math checks out on max reasoning.
In practice Opus still gets me further with less steering — it just understands the intent better.

Though with OpenAI’s resets right now, it’s still the better deal overall.

[Maurice @maurice_build]

"I get a lot more done with Claude"

NO, you don't. Unless you are using the 20x subscription.

With the current $100 subscription, you get the following monthly API equivalents:

Opus 5.5 | Claude Max 5x = $5,725
6.1 Sol | ChatGPT Pro = $1,055
Argon | Google AI Ultra = $300

If you compare the cost of completing a task at the max reasoning level, you will find that you get more tasks done with your GPT subscription:

Sol 6.1 | $1,055/ $0.72 ≈ 1,465
Opus 5.5 | $5,725/ $5.98 ≈ 957
Argon | $300 ÷ $1.99 ≈ 151

Even with high reasoning, GPT does beat Claude.

[Markizz @mar_kizzme]

Codex is especially bad at this — subagents there are mostly just extra noise
Sry

[Markizz @mar_kizzme]

Subagents are useless.
Without a clear, narrow purpose they just add overhead for a worse result than the main agent. The useful ones stay tightly scoped and run against the main flow; everything else mostly creates noise and zero-efficiency traffic between streams.

[Markizz @mar_kizzme]

I do so much work about work that I forget to actually develop

[Markizz @mar_kizzme]

@peduarte Thanks! Glad to be using it.

[Markizz @mar_kizzme]

Reinstalling Raycast was the best decision I made today.

[Markizz @mar_kizzme]

Looks like Claude’s week of generosity is over — no announcement, nothing. OpenAI with their resets is the better deal right now, even though I still like Opus 5.5 way more.

[Markizz @mar_kizzme]

Am I the only one whose Opus weekly limits went to $200 in a day? Feels worse than Codex.

[Markizz @mar_kizzme]

Dots → Codex orchestrator → Codex + Claude profiles
Try this setup. Seriously. It’s incredible.

[Markizz @mar_kizzme]

This is the most logical and appealing transition. I’ve always liked Grok—and Cursor. Now we get both of those—plus an X subscription—all rolled into one package. At this price point, it’s arguably the best deal out there right now; for just $20, it’s finally overtaken OpenAI.

[SpaceX Super Intelligence @SpaceXSI]

We've unified plans across Grok, Cursor, and X Premium.

Introducing The X Subscription.