Instagram story viewer> @rpn.tech> Posts

rpn.tech

@rpn.tech

104.4K

followers

4

following

Our Beautiful Dark Twisted Future
By @rpn

POSTS STORIES REELS TAGGED

Download All

When the client needs a crazy turnaround… you get the crew together in @slackhq

#Slack #SalesforcePartner by @rpn.tech

3k

157

3 days ago

The central dilemma of the intelligence age:

Moving at maximum speed could have devastating consequences.

Not moving at maximum speed could have devastating consequences.

Let’s dive into it.

Where do you stand? by @rpn.tech

5k

241

a month ago

OpenAI’s latest model, GPT-6 Astra represents a major step change in LLM capabilities.

It’s the most fun I’ve had using a computer in a long time. It feels like being a kid again, playing around with Photoshop, Macromedia Flash, or old school video game map/level editors.

Astra is currently the world’s most intelligent model, leading across major benchmarks:

98.6% on ARC-AGI-3
97.6% on FrontierMath Tier 4
72.6% on OSWorld 2.0 in about 47% less time
73.0% on DeepSWE

While this video is sponsored, everything I mention here is my honest assessment. I encourage you to give Astra your hardest challenge. Something you didn’t think was possible with AI. I think you’ll be surprised.

Follow me on Threads where I’m sharing more experiments.

@OpenAI #ChatGPT_Partner by @rpn.tech

9k

216

a month ago

OpenAI just announced GPT-6 Astra, calling it “a generational leap in capability.”

Co-founder Greg Brockman says “welcome to the AGI era.

Overall, GPT-6 Astra offers a breakthrough level of capability in coding, analytics, logic, and domain specific knowledge for dealing with complex enterprise knowledge work.

A few of the headline benchmarks:

- ARC-AGI-3: 98.6%
- Exploit-Bench: 100%
- FrontierMath Tier 4 (v2): 97.6%
- DeepSWE v1.1: 74.1%

According to OpenAI:

GPT-6 Astra is rolling out today to a limited set of organizations and over the coming days will become available to all ChatGPT Plus, Pro, Business, and Enterprise users, as well as through the OpenAI API and AWS. by @rpn.tech

2k

49

a month ago

Anthropic just launched Fable 5.1 and in short: it’s faster, smarter, cheaper… and finally responds to you like a sane, normal person would.

More specifics:

- Solves agentic coding tasks at roughly half the cost of Fable 5
- Much faster and slightly more capable than Opus 5, and uses around half of the tokens
- Writing style is more natural
- Much more powerful at scientific research

I chat with people in my day-to-day life all of the time who think AI is useless… and when I dig a little deeper I realize they are using base/commodity/legacy models.

So if you haven’t taken the time to seriously test-drive a frontier model over the last few months, this is a good opportunity to do so. They have gotten extremely proficient and I think it’s important for people to understand their capabilities.

The last few months specifically have been a real step-change, and I anticipate the next wave of models to close out the years will be another major step as well.

In this video I share some very basic things you can ask Fable 5.1 to do, so you can get a feel for how powerful it is. by @rpn.tech

8k

157

a month ago

Apple’s new AirPods with cameras have leaked! The macOS Tahoe 26.7 release candidate includes a demo video of AirPods with cameras built into thicker stems, designed for analyzing surroundings to power Apple Intelligence features like object recognition.

A few other details from the leak:

If you have long hair that blocks the camera, there will be an error that is titled "Hair Detected" that says "to get the most accurate information about things in your environment, make sure AirPods are not covered"

There is also something called Secure Pairing: "Please put your AirPods in case to continue Secure Pairing." by @rpn.tech

295

34

2 months ago

Anthropic just dropped Opus 5, which in their words is a “thoughtful and proactive model that comes close to the frontier intelligence of Claude Fable 5 at half the price.”

But the internet is shocked at how good it is at creating video games, CAD design, 3D experiences, etc.

This gives more credence to my theory that distribution is the final moat.

Also — pair Opus 5 with Claude Design — the early results in my tests have been incredible.

What has your experience been with Opus 5? by @rpn.tech

3k

355

3 months ago

Big news out of China: Moonshot AI has released Kimi K3, a 2.8 trillion parameter open-weight model with competitive benchmarks.

It is priced at $3 per million input tokens and $15 per million output tokens, and benchmarks show it outperforming Claude Opus 4.8 across multiple evaluations while ranking just behind GPT-5.6 Sol Max and Claude Fable 5 Max. by @rpn.tech

2k

30

3 months ago

Big news from @OpenAI - not only are they launching GPT 5.6 Sol, but they are merging Codex into the ChatGPT desktop app.

On desktop, there are now three ways to use ChatGPT: Quick Chat, Codex… and the newly introduced, ChatGPT Work.

Quick chat is the ChatGPT experience you’re already familiar with, just now in a sidecar. I found this experience very useful as I no longer had to switch tabs back and forth during a work session.

Codex is still there for developers, and improved greatly with the new underlying models.

And Work is a new agent that can take action across your apps and files, stay with a project for hours (and hours and hours) if needed, and turn a goal into finished work.

Let’s talk about GPT 5.6. I’ve had early access to it, and it is extremely powerful (and ~half the price of Fable)

To me, it is superior to any model out there in two key areas:

Writing: it is an incredible collaborator for producing content scripts, marketing emails, etc. It writes very clearly and cleverly, without any of the odd syntax and language we’re used to from past AI models.

UI Design: it seems to have a real design “intuition.” It makes surprisingly good choices in hierarchy, composition, motion etc that I haven’t yet seen in another model.

Have you had a chance to try it? Would love to hear how you like it. by @rpn.tech

10k

327

3 months ago

Scientists just read a 2,000 year old scroll… without opening it.

The scroll was part of a collection buried under ash and scorched into solid black lumps by the Mount Vesuvius eruption.

Now, without unrolling them, researchers were able to read one of them, uncovering what may be a work by a well-known stoic philosopher.

The breakthrough comes from the Vesuvius Challenge, an international research effort to digitally read the scrolls.

used a synchrotron to essentially X-ray inside the scrolls and detect the ink ancient Romans used to write. The letters are then studied by papyrologists, who translate the text.

Experts think the text reads more like a Stoic treatise on ethics and human behavior, and it specifically mentions Aristocreon, the nephew and pupil of the influential Stoic philosopher Chrysippus.

More than 6,000 scrolls, once owned by the father-in-law of Julius Caesar, remain unopened.

Follow @eluna.ai to stay updated with all of the latest and most important developments in the world of AI. by @rpn.tech

2k

46

3 months ago

Google Omni Flash is here and it rewrites the rules of marketing by turning what used to be a high-budget video production process into a simple, conversational workflow.

It’s available to try now on @higgsfield.ai

But you can actually interface with it through Claude, via the Higgsfield MCP.

Comment “OMNI” and I’ll send over my tutorial on how to do this.

#HiggsfieldCreator by @rpn.tech

22k

8k

3 months ago

Anthropic has announced that Fable 5 is officially back… but there is a caveat.

It has a one-week subscriber trial with a strict 50% usage ceiling.

Pro, Max, Team, and premium Enterprise users can use up to 50% of your weekly subscription limits on Claude Fable 5 at no extra cost.

After that, the free part of Fable 5 access ends.

Users can then pay with usage credits, or switch back to another Claude model. by @rpn.tech

1k

18

3 months ago