ClearText AI

AI obsessed

[ClearText AI @ClearTextAI]

@tankots @WisprFlow this corrected itself after the first 2-3 days of Duet. Maybe it was a Claude issue.

[ClearText AI @ClearTextAI]

@WisprFlow heads up, Wispr doesn't work whatsoever in the new Claude Duet (both app and web are completely non-functional for Wispr).

[ClearText AI @ClearTextAI]

@iScienceLuvr same. i've had an android since the original T-Mobile G1 lol NO REGRETS

[ClearText AI @ClearTextAI]

Talking to ChatGPT vs Talking to Google 😂

[ClearText AI @ClearTextAI]

@SamoBurja So the standard we hold our "leaders" to, who we've entrusted a large chunk of the US economy on falls back to

"well.....he was hardly alone"

Lol. Between Jensen and the prez, why is there reason for anyone to follow any rule?

[ClearText AI @ClearTextAI]

One more thing. I reran Claude Opus 5.5 with the exact settings from its launch day in September: same model name, same request, same thinking level.

Max still comes back empty. One level down, it’s a different animation than the one I got seven weeks ago. It now draws a gloved hand holding the bar, the balls cast shadows, and the gravity moved from a little floaty to a little fast. It also thought about two and a half times longer and cost twice as much.

Same name, different results. I can’t tell you why from one run each, but it’s worth knowing that the model behind a name can change.

Nobody has made the bar drive the puppet yet.

[ClearText AI @ClearTextAI]

The Marionette Benchmark: Fable 5.2 vs. Gemini 4 Carbon vs. Sonnet 5.5 vs. GPT-6 Astra

Weirdly, Sonnet 5.5 is still the best model in the world at this test. Each model writes the whole animation without ever seeing what it made. First up, the flagships.

Claude Fable 5.2 High | (15m 39s / $3.59)*

The best scene anyone has built. A city skyline behind the curtains, a briefcase sitting on the stage, the first prop we’ve seen. His eyes follow the balls, his mouth moves, his feet are on the floor and the balls cast shadows. Then you notice how fast they fall, about twice Earth gravity. Nothing holds the bar either. It looks like the bar is working his hands, but in the code it’s the other way around. @claudeai

Gemini 4 Carbon xHigh | (4m 02s / $0.28)*

The best staging of the group: a dark room, one spotlight, and dust floating in the light. Ball shadows, puppet shadow, and a cord running from the bar up out of frame, which almost nobody draws. The gravity is a touch fast and his feet hang well above the floor, which throws the whole thing off. @OfficialLoganK

GPT-6 Astra Max | (17m 42s / $2.03)
Still OpenAI’s flagship, from September. Moon gravity and a knee that bends backwards. @OpenAI

Claude Sonnet 5.5 xHigh | (13m 15s / $1.06)*

Still the leader. I put Fable 5.2 next to it and Sonnet won: close to Earth gravity, and a gloved hand actually holding the bar.

*Ran below the top setting. Fable 5.2 and Sonnet both came back empty on max. Fable burned $10.24 thinking before it produced anything.

[ClearText AI @ClearTextAI]

The Marionette Benchmark: Gemini 4 Argon vs. Muse Wonder vs. Sonnet 5.5 vs. GPT-6.1 Sol

Two new frontier models this week. Same test as always: a marionette only works if the bar above moves the limbs below, and juggling only works if the balls obey gravity and the hands take turns. Each model writes the whole animation without ever seeing what it made. This grid has the latest from each of the four big labs.

Gemini 4 Argon Medium | (1m 28s / $0.09)*
The idea is the most ambitious here. It builds the juggling in depth, so the balls pass in front of and behind the body instead of through each other. Then you look at the puppet. His hands come off the ends of his arms for half of every throw and float on their strings. He’s standing on the back wall instead of the stage floor, and his far shoe melts into his shadow. One sleeve has a white cuff and the other doesn’t. The eyes look like they’re following the balls, but they’re painted on and just nod along with the head. The balls drift down slowly, with no shadows.

Muse Wonder Max | (3m 05s / $0.52)
The first model where the balls fall too fast. Everything else we’ve tested juggles like it’s on the moon. This one looks like hypergravity. The string work is the liveliest here, with two strings running to an arm you can barely see behind his body, and his legs move with the act. The balls cast shadows and the stage has real lighting. Feet hover just off the floor, and the suit is boxy. Meta’s chart shows a “Full” version of Wonder, but that one isn’t available to developers yet, so this is the regular model.

Claude Sonnet 5.5 xHigh | (13m 15s / $1.06)*
Still the leader. A gloved hand holds the bar, the knees bounce with the throws, and the gravity is close to Earth.

GPT-6.1 Sol Max | (10m 12s / $0.32)
Still OpenAI’s best, and still floaty.

*Ran below the top setting. Sonnet came back empty on max. Argon’s top setting got cut off partway through by Google’s output limit.

nobody has made the bar drive the puppet yet.

[ClearText AI @ClearTextAI]

Small models

Gemini 3.8 Flash Dynamic | (3m 40s / $0.11)

The best of the small models, and it isn’t close. The arms really pump through every throw and catch, and the gravity is right around Earth. Bare stage, nice lighting, and the skinniest puppet with the smallest head we’ve seen. @GeminiApp

Muse Ember Max | (1m 09s / $0.02)

The first puppet that blinks. A cord runs from the bar up out of frame, and there’s a little nameplate on the stage that says “The Business Juggler.” The arms barely move, so the balls look like they’re juggling themselves, and they float like they’re in space. @AIatMeta

Claude Haiku 5.5 xHigh | (6m 44s / $0.05)*

Came back empty on max, like the rest of the Claude 5.5 family. Nine strings, the most we’ve seen, no stage at all, and two of the balls pass through each other.

GPT-6 Luna Max | (6m 32s / $0.02)

A clean side view, but floaty, and it throws higher going one way than coming back.

[The Midas Project @TheMidasProj]

🚨NEW: Dozens of accounts defending Jensen Huang after his affair with an alleged Chinese spy appear to be fake.

They are part of a much wider pro-Nvidia campaign, linked to an Israeli influence firm that says it shut down last year…

But the accounts never stopped posting. 🧵

[ClearText AI @ClearTextAI]

@iansilber will somebody check on Arfur Grok, this isn't like him

[ClearText AI @ClearTextAI]

@fidjissimo @ArfurGrok @ChronicleBioAI A rare L for ArfurGrok

[Chris McGuire @ChrisRMcGuire]

The story keeps getting crazier. To recap everything that has been reported so far:

- The US intelligence community discovered that Jensen Huang, the CEO of the largest company in the world, was sleeping with Alice Huang, a Chinese spy and an executive at Megaspeed. (FT)
- Multiple independent investigations have concluded that Megaspeed is likely a Chinese cutout, which helps Chinese companies access AI chips or circumvent export controls. (NYT and Bloomberg)
- The US intelligence community also concluded that Alice Huang’s goal was to convince Jensen to lobby the U.S. government to loosen export controls on advanced AI chips. (NYT)
- The intelligence community relayed this information to “a small group of top US government officials” earlier this year. (FT)
- Over the last year, Jensen Huang has repeatedly lobbied the U.S. government to loosen export controls on advanced chips. @BISgov has done so, not only by permitting the export of H200 chips to China, but also not enforcing export controls that would have required a license for exports of the most advanced Blackwell chips to Megaspeed and Southeast Asia.
- Although @CommerceGov leadership was presumably briefed on this information by the U.S. intelligence community, @BISgov has taken no actions to tighten controls on Megaspeed or Southeast Asia.
- Megaspeed remains a preferred cloud partner of Nvidia, and continues to receive Nvidia Blackwell chips despite employing a Chinese spy who reportedly is romantically engaged with the CEO of Nvidia.
- Tens of billions of dollars worth of AI chips have been exported to Southeast Asia this year alone, and Malaysia and Singapore are now two of the four largest importers of AI chips in the world.

What is going on here??? We need answers. The question of who gets the most advanced AI chips in the world is literally the defining strategic question of our time. And not only does it seem like the MSS may be directly influencing the debate on export controls in the United States, but also @CommerceGov is refusing to listen to the warnings of the IC on this issue. What is a POSSIBLE good explanation for why Megaspeed is still receiving chips? Why is there not a license requirement for AI chip shipments to all of Southeast Asia? This is beyond strange at this point.

The fallout from this story will be massive. Obviously, the motivations underpinning Nvidia’s policy views opposing export controls must now all be questioned. But also, how did @BISgov sit on this information and has still taken no actions? We need answers.

[ClearText AI @ClearTextAI]

@itmeolliev @nikitabier my brother in christ he worked for Elon, it is not a concern

[ClearText AI @ClearTextAI]

@AndrewCurran_ When you peel back all of the layers of this issue. The general population outside of the X AI bubble have serious concerns about AI's effects on society.

Those concerns are handwaved away here, not in the real world. These results will continue until they're taken seriously

[Miranda Nazzaro @mirandanazzaro]

From one Dem strategist tonight on AI industry’s losses tonight:

“It’s almost like ‘let the President and his AI billionaire friends do whatever they want’ wasn’t a winning message.”

[ClearText AI @ClearTextAI]

There's definitely an exodus of the Old Anthropic Guard happening.

[cat @_catwu]

Friday was my last day at Anthropic.

Two years ago, when Claude Code was a small internal tool, I asked it to build an RL env. I steered it for tens of hours but never manually wrote a line of code. It worked, I was hooked, and I joined the team.

When we first launched, most people approved every permission prompt by hand and mostly used Claude to answer questions and write tests. 

Today, the default is trusting Claude's judgment to auto-approve. Claude does large migrations and builds new products. 

Helping build Claude Code into an industry-defining product has been the journey of a lifetime. 

Thank you to the incredible team behind Claude Code, Cowork, and Tag. These products are the sum of thousands of great ideas from you. I'll miss working with you all.

Thank you to our users for embracing a new way of working. Your constant feedback shaped so much of what we shipped and your stories made the work so meaningful.

Excited to see what you all make next!

[ClearText AI @ClearTextAI]

Facts

[Ivan Fioravanti @ivanfioravanti]

Avoid any yearly prepaid LLM plan out there.

Labs are starting to reduce the quota to increase revenues.

Churning is your only weapon.

[ClearText AI @ClearTextAI]

@TimJayas Can you truly call a model that moved down only 2 slots after a recalibration "benchmaxxed"?

[ClearText AI @ClearTextAI]

@8teAPi @SenseNopedOut you can access duet from the claude desktop app, its hidden here