651–660 of 1001 posts
API pricing is up to $2/M for input and $12/M for output For comparison: Gemini 2.5 Pro was $1.25/M for input and $10/M for output Gemini 1.5 Pro was $1.25/M for input and $5/M for output
* 1,500 RPD (free), then $35 / 1,000 grounded prompts
to
* 1,500 RPD (free), then (Coming soon) $14 / 1,000 search queries
It looks like the pricing changed from per-prompt (previous models) to per-search (Gemini 3)
I love it that there's a "Read AI-generated summary" button on their post about their new AI. I can only expect that the next step is something like "Have your AI read our AI's auto-generated summary", and so forth until we are all the way at Douglas Adams's Electric Monk: > The Electric Monk was a labour-saving device, like a dishwasher or a video recorder. Dishwashers washed tedious dishes for you, thus saving you…
after outsource developer job, we can outsource all of manager job and leaving CEO with AI agentic code as its servant
Not a single engineer has ever been laid off because of AI. Any company claiming this is the case is trying to cover up bad decisions.
"Were automating with AI" sounds better to investors than "We over hired and now need to downsize" or "We made some bad market bets, now need to free up cash flow"
Earlier quoted context omitted.
Having seen the page the other day this is pretty incredible. Does this have the same 2000 token limit as the other page?
This isn't using the same prompt or stack as the page from that post the other day; on aistudio it builds a web app across a few different files. It's still fairly concise but I don't think it's that much so.
This blows Sonnet 4.5 (64.8%), GPT5.1 (58.7%) and Opus 4.1 (66.3%) away.
An example output:
'''
The cabin was cold. It smelled of pine needles and old dust. Bryan sat on the porch with a glass of scotch balanced on his knee. It was cheap scotch. He missed the good stuff, but he wasn't in Paris anymore, and he certainly wasn't in Beverly Hills.
He checked his watch. It was a bulky thing, black rubber. He checked the Glock on the table next to him. That was black, too.
"I told you," Bryan said. He wasn't talking to anyone. "I told you not to go into the woods."
He was thinking about Kim. He was always thinking about Kim. She was safe, probably. At a mall in Santa Monica. Or maybe a library. But it felt like she was gone. It always felt like she was gone.
The trees at the edge of the clearing shifted. There wasn't any wind. The air was heavy and humid, unusual for October. A bird stopped singing. Then the crickets stopped. It got very quiet.
Bryan took a sip of the scotch. It burned his throat. He liked the burn. It reminded him he was still there.
Three red dots appeared on his chest. They made a small triangle.
Bryan didn't look at them. He set the glass down on the railing. He picked up the phone. It was an old flip phone. He pressed a button.
"I don't know who you are," Bryan said into the phone. "I don't know what you want. If you are looking for ransom, I can tell you I don't have money."
The leaves rustled. A shape began to form out of the empty air. It looked like heat rising off the highway in August. It was tall. Much taller than a man.
"But what I do have," Bryan said, "are a very particular set of skills."
The shape solidified. It was ugly. It had skin like a lizard and dreadlocks that looked like heavy black cables. It wore a metal mask. It made a clicking sound. Click-click-click. Like a Geiger counter.
"Skills I have acquired over a very long career," Bryan said. He stood up. His knees popped. He felt the ache in his lower back. He was getting too old for this.
The thing roared. It wasn't a human sound. It was a wet, guttural noise.
"I will look for you," Bryan said. He put the phone in his pocket. "I will find you."
The creature raised a metal gauntlet. Two jagged blades slid out. They caught the moonlight.
"And I will kill you," Bryan said.
He raised the Glock and fired twice. Pop-pop.
The creature didn't fall. It took a step forward. Green blood, thick and glowing, leaked from its shoulder. It looked like antifreeze. The creature touched the wound, then looked at its hand. It seemed surprised.
Bryan didn't wait. He moved off the porch. He moved fast for a man his age. He went left, toward the woodpile. The creature fired a blast of blue light from its shoulder. It hit the porch swing. The swing exploded. Wood splinters rained down like confetti.
Bryan rolled behind the logs. He breathed in. He breathed out. He could hear the heavy footsteps crunching the dry leaves. Crunch. Crunch.
"You're big," Bryan whispered. "That makes you slow."
He grabbed a splitting maul from the stump. It was heavy. The handle was smooth hickory.
The creature came around the woodpile. It was scanning the trees. It didn't see Bryan down low. Bryan swung the maul. He put his hips into it. He swung it like he was chopping a cord of oak.
The blade hit the creature in the knee. There was a snap. A loud, wet snap.
The thing howled. It fell onto one knee.
Bryan dropped the maul. He stepped inside the creature’s reach. He knew exactly where to hit. The throat. The armpit. The soft spot under the ribs. He hit the creature three times, hard chops with the side of his hand. It was like hitting a brick wall, but he felt something give.
The creature swung its arm back. It hit Bryan in the chest.
Bryan flew backward. He hit the dirt. The wind went out of him. He lay there for a second, staring up at the stars. They looked very far away. He wondered if Lenore was looking at the same stars. Probably not. She was probably sleeping.
He sat up. His ribs hurt. Maybe broken.
The creature was trying to stand. It was clicking again. It tapped something on its wrist. A series of red symbols started flashing. They counted down.
Bryan knew a bomb when he saw one.
"No," Bryan said.
He tackled the thing. He didn't think about it. He just did it. He grabbed the creature’s arm. He twisted the wrist mechanism. He’d seen something like it in Baghdad once. Or maybe Istanbul. The memories ran together now.
He ripped the gauntlet loose. Wires sparked. He threw it as hard as he could into the darkness of the woods.
Three seconds later, there was a flash. A boom. A shockwave that shook the pine needles from the trees.
Silence came back.
The creature lay on the ground. It was breathing shallowly. The green blood was pooling under it. It took off its mask.
The face was hideous. Mandibles. Beady eyes. It looked at Bryan. It said something, a garbled copy of Bryan's own voice.
"...good luck..."
Then it died. It just stopped.
Bryan stood up. He dusted off his pants. He walked back to the porch. The swing was gone. The railing was scorched.
His glass of scotch was still sitting there, untouched. The ice hadn't even melted.
He picked it up. He took a drink. It still tasted cheap.
He took his phone out and looked at it. No service.
"Well," he said.
He went inside the cabin and locked the door. He sat on the couch and waited for the sun to come up. He hoped Kim would call. He really hoped she would call.
'''
Earlier quoted context omitted.
Yeah, it is often pointed out as a brilliance in game analysis if a GM makes a move that an engine says is bad and turns out to be good. However, it only happens in very specific positions.
Does that happen because the player understands some tendency of their opponent that will cause them to not play optimally? Or is it genuinely some flaw in the machine’s analysis?
But mostly what happens is that a move isn't so good, but it isn't so bad either, and as the computer will tell you it is sub-optimal, a human won't be able to refute it in finite time and his practical (as opposed to theoretical) chances are reduced. One great recent example of that is Pentala Harikrishna's recent queen sacrifice in the world cup, amazing conception of a move that the computer say is borderline incorrect, but leads to such complications and a very uncomfortable position for his opponent that it was practically a great choice.
Earlier quoted context omitted.
These prediction markets are so ripe for abuse it's unbelievable. People need to realize there are real people on the other side of these bets. Brian Armstong, CEO of Coinbase intentionally altered the outcome of a bet by randomly stating "Bitcoin, Ethereum, blockchain, staking, Web3" at the end of an earnings call. These types of bets shouldn't be allowed.
The point of prediction markets isn't to be fair. They are not the stock market. The point of prediction markets is to predict. They provide a monetary incentive for people who are good at predicting stuff. Whether that's due to luck, analysis, insider knowledge, or the ability to influence the result is irrelevant. If you don't want to participate in an unfair market, don't participate in prediction markets.
A nice Easter egg in the Gemini 3 docs [1]: If you are transferring a conversation trace from another model, ... to bypass strict validation in these specific scenarios, populate the field with this specific dummy string: "thoughtSignature": "context_engineering_is_the_way_to_go" [1] https://ai.google.dev/gemini-api/docs/gemini-3?thinking=high...
It's an artifact of the problem that they don't show you the reasoning output but need it for further messages so they save each api conversation on their side and give you a reference number. It sucks from a GDPR compliance perspective as well as in terms of transparent pricing as you have no way to control reasoning trace length (which is billed at the much higher output rate) other than switching between low/high…
I love it that there's a "Read AI-generated summary" button on their post about their new AI. I can only expect that the next step is something like "Have your AI read our AI's auto-generated summary", and so forth until we are all the way at Douglas Adams's Electric Monk: > The Electric Monk was a labour-saving device, like a dishwasher or a video recorder. Dishwashers washed tedious dishes for you, thus saving you…
I am personally impressed by the continued improvement in ARC-AGI-2, where Gemini 3 got 31.1% (vs ChatGPT 5.1's 17.6%). To me this is the kind of problem that does not lend itself well to LLMs - many of the puzzles test the kind of thing that humans intuit because of millions of years of evolution, but these concepts do not necessarily appear in written form (or when they do, it's not clear how they connect to specif…
What I would do if I was in the position of a large company in this space is to arrange an internal team to create an ARC replica, covering very similar puzzles and use that as part of the training. Ultimately, most benchmarks can be gamed and their real utility is thus short-lived. But I think this is also fair to use any means to beat it.
We have a global RL Pipeline on our hand.
If there is something new a LLM/AI model can't solve today, plenty of humans can't either.
But tomorrow every LLM/AI model can solve it and again plent of humans still can't.
Even if AGI is just the sum of companies adding more and more trainingdata, as long as this learning pipeline becomes faster and easier to train with new scenarios, that will start to bleed out humans in the loop.