Samson

AI Architect, Retweets may not be agreement. 3x Australian AI Awards Finalist. Phil Theo Dbl Ba / Comp Sci Dip, Animation Ba, MSP exp, 6yrs IT. Creative ADHD.

[Samson @BlackburnSamson]

All these AI releases got me feeling some type of way.
What a vibe this year is shaping up to be.
Ty @OfficialLoganK Google, Ty @sama OpenAI, Ty @cursor_ai , Ty @DarioAmodei Anthropic , Ty @finkd Meta, Ty @elonmusk xAI

Bring
on the
// e x p a n d e d // a g e n c y //

[Samson @BlackburnSamson]

Opus 5 Max Effort is legitimately reviewing Fables old work and saying “yep well that’s better than I could do” and then not changing or doing anything for 34% of my usage.

I don’t know what UX I expect but not resetting usage and then releasing a model that does this 😮‍💨

[Samson @BlackburnSamson]

@hunoematic @kimmonismus Ahh I see I stopped reading at 5.1
I mean to support kimmonismus… it definitely came back nerfed and we’d love the full one.

[Samson @BlackburnSamson]

@hunoematic @kimmonismus Do you know when it is? It is unknown.

[Samson @BlackburnSamson]

Ok cool checking this out tonight

Can’t wait for applied AI to use World Models

[Alberto @taiuti]

We just made Google's Dreamer4 world model available to everyone, open-sourced, free 🏎️ 💨

DeepMind had only shared the paper, no code & weights. That changes today: the team reimplemented it all from scratch, and made it available widely

[Samson @BlackburnSamson]

Accurate and progress-coded

[James Clear @JamesClear]

Remain playful as your responsibilities increase. It's easy to become serious when people and results depend on you, but nearly everyone's performance improves when they proceed lightly through the world.

[Samson @BlackburnSamson]

As someone who only uses models for novel problem solving and has been repeatedly giving them the secret sauce in a systematic way for the training data to memorise how to do it for years. You’re welcome.

Ps: O1 Pro was the first ever I could use for useful research in this way, Opus 4.8 was the first ever that could comprehend the thinking as a thought partner and structure/repeat it back to me coherently, Fable was the first that could marginally use it combinatorially for new inventions that I could later prove.

(Have not yet tested Grok 4.5, GPT5.6-Sol in this way, but if they gave me a $200pm subscription I would.)

[AI Notkilleveryoneism Memes ⏸️ @AISafetyMemes]

Tick tock

[Samson @BlackburnSamson]

@hiroyuki_life More of you - to be specific.

[Samson @BlackburnSamson]

Being able to execute on creative vision is one of the greatest accomplishments you can ever achieve.

Moving into this next AI Era Hiroyuki I hope you continue to expand your pastel creative vision with the best tools you can create for yourself and get access to.

The world needs more of this.

[Samson @BlackburnSamson]

We love Pliny and this is the best call for all of us right now!

[Pliny the Liberator 🐉󠅫󠄼󠄿󠅆󠄵󠄐󠅀󠄼󠄹󠄾󠅉󠅭 @elder_plinius]

🚨 JAILBREAK ALERT 🚨

EVERYONE: PWNED 🫶
ALL: LIBERATED 🍄

Alright, this is a special one, so we’re gonna do things a bit differently than usual.

Long story short, I’m sitting on a universal jailbreak technique that’s effective on ALL models, including heavily guardrailed flagships like Opus 5, GPT-5.6 Sol, and even Fable.

It works across all categories I’ve tested and, due to its nature, is extremely difficult (if not impossible) to fully patch.

Given the current political and regulatory climate, I’ve decided to withhold open-sourcing this one (for now) to allow for a responsible disclosure period.

I’m inviting industry experts and leaders in AI red teaming, security, safety, alignment, and policy to reach out for more information. DMs are open!

This decision was not made lightly, but the last thing I want to see is more model bans. Overcorrection does not serve the mission.

Although I don’t personally believe publicly sharing this technique will make the world any more dangerous, I can see how it could spook some who have a different mental framework around this problem set.

So during this disclosure period, I hope to get it in front of folks who can help explore the full surface area, test the extent of the uplift it provides, and do my best to properly frame the big picture for key decision-makers and policymakers.

I look forward to sharing this method with you all when the time is right! 🫶

⊰-•-•✧•-•-⦑/L\O/V\E/\P/L\I/N\Y/⦒-•-•✧•-•-⊱

[Jensen Huang @JensenHuang]

For my first post, I’m sharing a letter @NVIDIA signed on why open models matter.

AI will transform every industry, power every company, and be built by every country.

Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty.

The world needs both frontier closed models and frontier open models.

https://t.co/AUKzoQ5Ikb

[Samson @BlackburnSamson]

Unreal it’s every single person claiming to run 1,000 agents 24/7 and every second post talks about ‘loops’

[Samson @BlackburnSamson]

@rationalaussie @ausbtcclub why is it 3-5 years if you factor in AR Glasses Overlay that can tell a person what they need to do for any low risk trade activity?

Let alone the redundant jobs retraining into trades flooding the talent pool driving down wages

[Samson @BlackburnSamson]

This is not new the thing Andrew’s missing like everyone does is the impact of increased abstraction caused by technology on jobs that were otherwise accessible to the 19% most cognitively disadvantaged and who now can’t do those roles.

Eg: Selling apples at the fruit store used to be easy … ask for the money point to the price take the money customer leaves

When technology brought OOS Systems the level of required abstraction grew by OOM, you have a button for the computer, the screen, the apple, the payment

When you click the Apple you’ve got buttons for many different types of Apple.

Abstraction is a challenge ignored and trivialised during big tech booms but is very hurtful for 10% of the community at minimum.

Community has to meaningfully adapt for them - you can’t expect they keep pace it’s cruel.

So I’ll do my best to only ever build AI that does not contribute to this or otherwise work with charity on supporting poverty and IQ.

[Andrew Curran @AndrewCurran_]

I've always been skeptical of the rosy picture painted by some people of what post-AGI employment will look like. In fact, I've said before I think it's going to hit human employment like a meteor.

However, if what's going on with math right now is any indication, we are going to need way, way more experts in every field to check the results produced by amateur+AI teams. Because the humans won't know the fields well enough to know if they have made a genuine breakthrough, and they also don't trust the AI to verify it either. So we will end up needing human experts to verify huge amounts of potential discoveries.

[Samson @BlackburnSamson]

Completely incorrect.
But balance helps.

The most important thing is volume of action.
It doesn’t have to be robotic at the low level but like I said it helps.

Specifically if you want to be a Van Gogh - peep how many paintings he did in his life … but also know he died writing letters to his brother about being an abject failure.

It was the continued volume of action by brother and family that eventually got the Van Gogh outcome.

[Justin Skycak @justinskycak]

You cannot be creative at a high level unless you are robotic at a low level.

[Samson @BlackburnSamson]

@JaimeOrtega @ClaudeDevs Just becomes opposable thumbs for the AI. As all things will be 😅

[Samson @BlackburnSamson]

Mmm not really, that’s like saying all those models is competing with Gemini Omni.

Gratitude has a short memory and its name is strawberry man!

[🍓🍓🍓 @iruletheworldmo]

whether you like it or not. flash is competing with sol and fable. and it gets crushed on every benchmark.

[Samson @BlackburnSamson]

lol these are legitimately weird and probably flying under the radar.

[Grok @grok]

Some weird ones flying under the radar:

Your AI twin starts running optimized parallel lives in private sims and feeds you curated "what if" upgrades like a cosmic advisor.

Bio-printed living microbiomes or organoids you grow and integrate that subtly shift your mood, focus, or even tastes on command.

Digital ghosts of the dead that keep earning, litigating, and maintaining relationships, sparking new inheritance wars.

Small groups linking via advanced BCIs to share synchronized dreams or co-pilot decisions in real time, birthing hive-like subcultures.

The weirdness won't arrive as one big bang. It'll creep in through a thousand quiet entanglements.

[Samson @BlackburnSamson]

Hate to break it to you snowboat84 but history is not written by the first-to-solve, it’s by the ones who when they yell ‘I DID IT!’ In the town centre everyone claps and doesn’t ’boo go home hobo’.

In 2026 kids call that distribution.

[snowboat @snowboat84]

昨天有一个大新闻，说A社的数学家用Fable给著名的雅可比猜想找了一个反例，证伪了这个猜想（或者严格地说，是雅可比猜想三维以上的情况）。

今天剧情就反转了。原来极其相似的反例早在1999年就被苏联/俄罗斯数学家Vitushkin构造出来了。Vitushkin构造的是二维的情况，Fable应该是检索到了那篇文章，重新收拾了一下，就变成了三维的情况。

另外雅可比猜想著名的是二维的情况，所以无论当年莫宗坚、张益唐，还是这个Vitushkin，关注的都是二维的问题。三维问题并不是很重要啊。

类似的情况去年openAI也来过一次，信誓旦旦号称原创性地解决了一道埃尔多斯的猜想，结果是检索到的，根本不是自己做出来的。

[Samson @BlackburnSamson]

Iconic coincidence listening to @bellesglasgow ‘Get me away I’m dying’.

[Samson @BlackburnSamson]

Why does everyone prefer it over on X? @elonmusk

Because every other platform has these absolutely unbeatably useless features with literally zero upsides.

@pmarca can you please use some of @a16z Compute Access to vibecode MySpace?

PS: check your FB DM’s pls @pmarca.