10114
Discover the future of AI with ChatGPT! 🤖💬 Get AI insights, tips, tricks, and more. Questions: @AI_inquires_bot
The pattern is obvious: the stack is shifting from answers to action.
Читать полностью…
This week AI stopped looking like a chatbot race and started looking like a workflow race.
Читать полностью…
The sudden shift suggests the rivalry between xAI and Anthropic may be turning into something more cooperative.
Читать полностью…
He praised Anthropic’s models, values, and decision to keep advertising off its platform.
The change comes weeks after Anthropic announced a partnership giving it access to SpaceX computing power.
It can build websites
It can make interactive visual explanations
It can create polished presentations
And handle complex documents or financial models with more precision
GPT-5.6 also powers Codex with stronger coding performance, fewer tokens, and better long-horizon workflows.
Читать полностью…
Claude found a pricing loophole
A developer built an open-source proxy called pxpipe that can cut Claude API bills by turning bulky text context into PNG images.
The idea works because image tokens are priced by pixel size, not by how much readable text is inside the image. According to the project, a dense 1928 by 1928 image can hold around 92,000 characters while costing about 4,761 vision tokens.
That claim matters because math discovery is one of the hardest tests for AI. It means the model is not only answering questions, it may be helping researchers find ideas that were not known before.
Читать полностью…
The Credibility Test
“You are a reader who immediately distrusts writing that sounds too clean. Go sentence by sentence and rewrite anything that feels constructed, overcorrected, or too precise. It should read like something a real person said out loud without second guessing every word.”
Thought Flow Fix
“This is not a grammar fix. It is a thinking fix. Rewrite this so the ideas move the way a real mind moves, uneven in places, punchy in others, sometimes slower. Break any pattern where the writing feels too controlled or too evenly paced.”
Pattern Disruptor
“You specialize in spotting the tells that expose AI writing. Go through this and strip them out one by one. Vary the structure, soften rigid phrasing, and make each word feel chosen in the moment, not generated by a pattern.”
People can tell when writing comes from ChatGPT or Claude.
Not because the ideas are bad, but because the writing often feels too smooth, too balanced, too polished, and weirdly empty.
These prompts are built to fix that.
They push the model away from perfect sentences and toward writing that sounds like it came from someone with an actual voice, actual opinions, and actual friction in the way they think.
Save these for the next time your draft sounds like AI wearing a human costume.
Which AI models cost the most to get the job done?
This chart ranks models by the weighted average cost to complete one Artificial Analysis Intelligence Index task.
In simple terms: it shows the real cost per task, not just the price per token.
Most models were tested at their highest reasoning setting.
The smartest model is not always the best choice.
What matters is finding the cheapest model that can reliably do the job.
The model is still in training, and Wang said the claim is based on major AI benchmarks. Meta has not publicly shared which benchmarks were used, so the claim still needs outside proof.
Читать полностью…
Agents are the hot topic now. We are on the 5th layer now, iykyk
Читать полностью…
OpenAI, Microsoft, Google, Meta, xAI, and Anthropic all pushed harder on agents, voice, and multimodal tools, while robotics kept moving closer to real-world use.
Читать полностью…
🥺 Apple sues OpenAI over alleged AI hardware trade secrets
Apple has filed a lawsuit against OpenAI, accusing the company of orchestrating a campaign to obtain confidential information about Apple’s unreleased hardware projects.
The lawsuit names OpenAI hardware chief Tang Tan, Apple’s former VP of product design, along with former Apple engineer Chang Liu. Apple alleges Liu downloaded dozens of confidential hardware files before leaving the company and claims OpenAI encouraged departing Apple employees to share product designs, technical documents, and other proprietary information.
According to the filing, more than 400 former Apple employees now work at OpenAI, a figure Apple cites as part of its case.
Apple is asking the court to order OpenAI to destroy any confidential materials it obtained and redesign any upcoming hardware products that rely on Apple’s alleged trade secrets.
Musk has also recently highlighted Claude’s coding performance, while other xAI leaders have praised Anthropic’s work.
Читать полностью…
Elon Musk just reversed course on Anthropic.
After repeatedly criticizing the company and calling it “Misanthropic,” Musk now says he was wrong.
But the bigger update might be ChatGPT Work.
Instead of just answering questions, ChatGPT can now turn a goal into finished work
OpenAI just launched GPT-5.6, its new model family built for coding, knowledge work, cybersecurity, and complex agent tasks.
Читать полностью…
That creates a strange arbitrage. In one production snapshot, pxpipe says a bill dropped from $100 to about $41 by converting older context, tool docs, and system prompts into images before sending them to Claude.
The catch is accuracy. The compression is lossy, and the failure mode can be silent. Fable 5 recalled exact 12-character hex strings correctly in 13 of 15 tests, while Opus failed all 15 in the same dense render test.
This makes pxpipe useful for bulky context, but risky for IDs, hashes, secrets, or anything that needs exact recall.
If this holds up, GPT 5.6 could be one of OpenAI’s most important releases yet.
Читать полностью…
GPT 5.6 is here
OpenAI’s next wave of models is reportedly live, with Sam Altman saying GPT 5.6 can now help discover new math. The models are named Sol, Luna and Terra.
Voice Shaper
“You are building a voice that sounds like it belongs to one real person. Rewrite this so it carries a clear point of view, not the tone of a neutral observer trying to please everyone. Let opinions, small contradictions, and natural shifts in tone come through.”
Make It Hit
“Rewrite this with one goal, it should land when someone reads it. Keep what matters, but cut anything that feels flat, calculated, or forgettable. It should sound like it came from someone who actually cares.”
The Real Person Rewrite
“You are an editor with a sharp eye for writing that reads as constructed, not spoken. Rewrite this the way someone who actually lived it would say it, a little rougher, more direct. Cut anything that feels rehearsed, overbuilt, or written to sound impressive.
Anthropic’s leak got messy
A viral post claimed an Anthropic engineer leaked the company’s internal “Obsidian Brain” and was fired the same day.
But the confirmed story is different. Anthropic said a Claude Code release accidentally included internal source code because of human error during publishing.
Claude Code is Anthropic’s AI coding tool. The leak reportedly exposed parts of how the product works, including internal code and memory-related systems, but Anthropic said no customer data or credentials were exposed.
Boris Cherny, creator of Claude Code, said the focus should be on improving the process, not blaming one person.
The real story may be less dramatic than the rumor, but it still shows how easily internal AI systems can slip into public view.
Watermelon reportedly uses around 10x more compute than Avocado, Meta’s previous internal model. Wang also hinted that a stronger Muse Spark update is coming soon, with major improvements in coding and agentic abilities.
If this is accurate, Meta’s huge AI spending and talent push may finally be paying off.
Meta says it caught up
Meta’s superintelligence chief Alexandr Wang reportedly told employees that the company’s next AI model, codenamed Watermelon, has caught up to OpenAI’s GPT-5.5.