Atlas Can

Serial Entrepreneur & AI Expert | Owner of Inverta | Building the Future of AI OS for Agencies & Enterprise Businesses

[Atlas Can @itsatlascan]

I never thought Opus 4.8 would feel cheap until I tested Fable 5.

One hour on the new @claudeai Mythos-class model and my entire Claude 20x 5-hour limit was gone.

Honest review: the model is genuinely good definitely the best LLM out there right now.

But the flaw is the price. It's included in all plans until June 22 but after that it moves to usage credits, and this model is going to cost you lots.

So I'm not handing a full coding session to it.

My current setup: planning on Fable 5, execution on Opus 4.8. Saving on cost and pull more value out of both.

[Atlas Can @itsatlascan]

@claudeai Finally Months of frustration, cancelled subscriptions, switching to Codex.. and then THIS drop. 🎉

Fable 5 is Mythos-class and it’s in our hands right now. @AnthropicAI you made us wait but you delivered. Can’t wait to build with new model 💪🏼

[Atlas Can @itsatlascan]

@higgsfield @higgsfield_ai I have no idea what @higgsfield is up to, doesn’t feel like a needed update by them but not a big gamer so want to know what all gamers think about this ?

[Atlas Can @itsatlascan]

@JeffBezos last night. Most accurate Succession IRL 🤣

This event made me respect @SpaceX and @elonmusk even more 👏

[Atlas Can @itsatlascan]

@Polymarket They should have just invested in to @claudeai as equity 🤣

[Atlas Can @itsatlascan]

Claude Opus 4.8 is out. Finally New Model From @AnthropicAI .

But the real question isn't whether it beats GPT-5.5 on a chart. It's whether 4.8 actually closes the gap Opus 4.7 left open — or it's a filler before Mythos lands so Anthropic doesn't lose the race.

What changed from 4.7 to 4.8:

- Agentic coding (SWE-Bench Pro): 64.3% → 69.2%
- Reasoning with tools (HLE): 54.7% → 57.9%
- Computer use (OSWorld): 82.8% → 83.4%
- Fast mode runs 2.5x quicker at a third of the cost
- Roughly 4x less likely to leave silent flaws in code it wrote
- Fixes the tool-calling and comment-verbosity issues 4.7 shipped with

Anthropic is calling it a modest but tangible upgrade. Honest framing. The real prize is still Mythos, which they've said is coming in the next few weeks once the cybersecurity safeguards are done.

So my read: 4.8 is the bridge before Mythos. Better judgment in Claude Code, cleaner tool use, cheaper fast mode. Not the model that wins the race, but the one that keeps them in it until Mythos ships.

Anyone built on 4.8 in Claude Code yet? Used it on a real run? Noticed a real jump from 4.7?

I'm going to put it on a live build tonight and post an honest review tomorrow.

[Claude @claudeai]

Introducing Claude Opus 4.8: it builds on Opus 4.7 with sharper judgment, more honesty about its own progress, and the ability to work independently for longer than its predecessors.

Available today at the same price.

[Atlas Can @itsatlascan]

@claudeai Finally a new model from @claudeai — let's see if it actually holds up, and wins against GPT 5.5 or if we get the same story Opus 4.7 gave us.

Either way, still waiting on Mythos. The cybersecurity angle makes sense, but the wait is getting long.

[Atlas Can @itsatlascan]

@OpenAI Live translation across 70+ languages in real time.. 🤯 This completely changes how the world communicates. Receptionists at tourist spots like ( Turkey & Japan, etc.. ) , political summits, international business — no more language barriers, no more waiting.

[Atlas Can @itsatlascan]

Do you understand what @OpenAI just shipped..

1️⃣ GPT-Realtime-2 — GPT-5-class reasoning in live voice. Listens, thinks, and responds in a single audio loop with zero delays or silence.

2️⃣ GPT-Realtime-Translate — live translation across 70+ languages into 13 output languages, keeping pace with the speaker in real time.

Think about what this actually means. Tourist destinations like Turkey, Dubai, Japan — multilingual AI receptions at a fraction of the cost.

Political summits where human translators charge thousands per day — now at $0.034 per minute. Live customer support in any language with no hold times.

The translator profession, multilingual call centres, voice receptionists — all of it is about to look very different.

This isn’t the future. It’s already here about to replace all receptionist and translators.

[OpenAI @OpenAI]

Introducing GPT-Realtime-2 in the API: our most intelligent voice model yet, bringing GPT-5-class reasoning to voice agents.

Voice agents are now real-time collaborators that can listen, reason, and solve complex problems as conversations unfold.

Now available in the API alongside streaming models GPT-Realtime-Translate and GPT-Realtime-Whisper — a new set of audio capabilities for the next generation of voice interfaces.

[Atlas Can @itsatlascan]

@Polymarket A few hours ago @SpaceX powered up @claudeai on the ground with Colossus 1.

Now Anthropic wants to move the data centers to space for gigawatt-scale compute.

The compute race is about to leave the planet. 🚀

[Atlas Can @itsatlascan]

Big day for @claudeai — doubled usage limits, SpaceX compute deal, and now Claude literally dreams between sessions.

It’s reviewing its own work, finding patterns, pruning outdated memory and self-improving for your next session. This is what real agent memory looks like.

@AnthropicAI is moving fast today.

[Atlas Can @itsatlascan]

Finally.. no more hitting walls mid-session. This is all we asked for @claudeai 🎉

One thing left — ship a more powerful model then Opus 4.7 maybe ( Mythos ) and we are back on top at the AI race and I believe you will @AnthropicAI

[Claude @claudeai]

Effective today, we are:

1) Doubling Claude Code’s 5-hour rate limits for Pro, Max, and Team plans;

2) Removing the peak hours limit reduction on Claude Code for Pro and Max plans; and

3) Substantially raising our API rate limits for Opus models.

[Atlas Can @itsatlascan]

First @cursor_ai got a supercomputer behind it — now @claudeai.

@SpaceX is becoming the compute backbone of the entire AI coding race. More compute means higher limits on Claude Code and the API — and that’s exactly where Anthropic needs to compete with Codex. No more hitting walls mid-session.

Anthropic needs to ship Mythos or something even more powerful now. The compute is there — no more excuses. Time to go back to the top.

Congrats @AnthropicAI and @SpaceX — can’t wait to see the new improvements.

[Atlas Can @itsatlascan]

@Polymarket The worst thing that can happen for vibecoder in mid session 😭

[Atlas Can @itsatlascan]

@Polymarket

[Atlas Can @itsatlascan]

I "was" a heavy Claude Code user — but today I'm officially cancelling my Claude Max 20X subscription.

This whole weekend I've been building software and my plan has been maxing out in half a session while doing literally nothing useful. It keeps missing entire sections, ignoring orders even with a detailed Progress.md file for each phase, and doing whatever it wants regardless of the instructions.

Opus 4.7 feels like Anthropic deliberately made it worse. I've given it every chance and it keeps failing on the basics.

What's heartbreaking isn't just the bad experience — it's that Claude keeps shipping new features and updates while the model powering it gets worse with every release. You can't build on a regressing foundation.

From today I'm officially moving to Codex. Done — until Anthropic stops cutting corners and starts shipping powerful models again.

@claudeai -> @OpenAI

[Atlas Can @itsatlascan]

Be honest — which AI model is actually the best for coding right now?

[Atlas Can @itsatlascan]

This week in AI was genuinely insane. 🤯

GPT-5.5 dropped and beat Opus 4.7. DeepSeek V4 became the biggest open-source model ever. Kimi K2.6 launched Agent Swarm. Google invested $40B into Anthropic. Vercel and Lovable got breached. Claude Mythos leaked on Discord.

And it's only April. This year is moving fast.

[Atlas Can @itsatlascan]

After a huge week in AI — what's your go-to model right now?

GPT-5.5 / Kimi K2.6 / Opus 4.7 / DeepSeek V4