AI & TECH

NEWS • TIPS • RESOURCES

[AI & TECH @AiWithBDN]

If they actually pulled that off in that timeframe — that's genuinely impressive.

One correction though: Fable 5 was taken offline by the White House due to security concerns and didn't become publicly available again until July 1.

Moonshot released Kimi K3 on July 16.

That leaves a 15-day window

[Zynex @0xzynex]

One guy built Kimi K3 alone and made over $10,000,000 in a single month. Then he went on stage and told everyone exactly how he did it. Almost nobody pressed play.

The talk sits on YouTube, one engineer walking through the machinery under Kimi, the open models that quietly closed China's six-month gap to zero.

Minute 5 kills the data wall, he shows how a new optimizer turns 50 trillion tokens into the equivalent of 100 trillion, calls it "almost like magic," and means it.

Minute 17 is the part US labs would never publish, a linear attention design that beats full attention across short and long context while staying cheap at a million tokens.

Minute 23 is the swarm, reward functions for orchestrating up to 1,000 agents in parallel, including a term that catches sub-agents faking work they never finished.

Minute 25 he puts up a training curve, 30 trillion tokens with not a single loss spike, and calls it the most beautiful curve of his life.

Consultants charge $500 an hour to guess at this. The primary source is free and timestamped.

Everyone's arguing about the model. The blueprint has almost no views.

[AI & TECH @AiWithBDN]

@K9Aasim The competition with Elon Musk is now a direct face-off. It will be a net positive for the world.

[AI & TECH @AiWithBDN]

🚀 HISTORIC: China just achieved its first ever successful controlled first stage recovery of a carrier rocket

The Long March 10B lifted off from the Hainan Commercial Space Launch Site moments ago.

Roughly 6 minutes after stage separation, the first stage performed a vertical return and successfully landed on an offshore recovery platform.

This marks not one, but two global milestones in one launch:
✅ China's first ever successful launch vehicle first stage recovery
✅ The world's first ever net-capture recovery of an orbital carrier rocket

Watch the full on-site landing footage below 👇

#ChinaSpace #RocketRecovery #LongMarch10B #CommercialSpace

[AI & TECH @AiWithBDN]

There's a widely-circulated observation in China: the top leadership really likes Elon Musk. They think whatever he's doing is the right direction.

So when Musk went into electric vehicles, China followed. When he went deep into AI, China followed.

When he started building rockets? An entire private space industry sprung up — companies like CAS Space. When he pioneered rocket recovery? They're developing an entire zone in the South China Sea to build out that exact capability.

The most effective industrial policy blueprint in China right now might just be whatever Elon does next.

#China #ElonMusk #IndustrialPolicy

[Beff (e/acc) @beffjezos]

It really is China vs Elon on all fronts

[AI & TECH @AiWithBDN]

One of the most underreported shifts happening in AI right now: U.S. companies are adopting Chinese AI models at a staggering rate 📊

Data from OpenRouter, the platform that lets you switch between every major model, tells the story:
Token usage from Chinese AI models by U.S. businesses crossed 30% in February and peaked at 46%. That's up from just ~4% in the first half of 2025.

The math is impossible to ignore:
Anthropic's Claude Opus 4.7 costs $25 per 1 million tokens.
DeepSeek V4 Flash, the most popular Chinese model, costs $0.18.
That's less than 1% of the cost.

U.S. AI startup Lindy already made the switch, moving their email and meeting note generation service from Anthropic to DeepSeek. Their CEO said performance is identical and they'll save millions of dollars in just months.

There's a second factor driving this too: U.S. export controls. When Anthropic was ordered to restrict overseas access to Mythos and Fable in June, it created a market consensus that you can't guarantee long-term stable access to U.S. AI models.

Chinese AI models are mostly open-source, downloadable to your own servers — the sudden cut-off risk is minimal.

It's the classic enterprise tradeoff: massive cost savings on one side, geopolitical risk on the other. And right now, for more and more companies, the math is winning.
Source: Nikkei
#AI #OpenRouter #DeepSeek

[AI & TECH @AiWithBDN]

If you zoom out, you realize July 2026 isn't just a bunch of model releases — it's the entire AI industry simultaneously dropping its price cards 💰

Look at how every single launch is strategically positioned to undercut someone else:
- Fable 5 sets the new capability ceiling
- Opus 5 immediately takes that same capability tier and slashes the price
- Grok 4.5 undercuts Opus even further
- GPT 5.6 comes in cheaper than all of them, faster, and still competitive on capability
- Meta's Muse Spark 1.1 is basically saying "we'll give you 90% for free"
- China's Seedream 5.0 Pro is already eating OpenAI's lunch on commercial design use cases

This isn't just technical progress. This is coordinated price discovery happening in real time across the entire industry.

Six months ago, frontier capability cost $20+/month. Three months ago, $10.

Right now? Everyone's racing to see who can get it to $5 first.

The winner isn't going to be whoever builds the smartest model. It's going to be whoever can build a smart enough model — at a price point no one else can match.

#AI #PriceWar #GPT5.6

[AI & TECH @AiWithBDN]

📸 ByteDance's flagship image model just went head to head against OpenAI's latest — here's how it stacked up

I ran the exact same prompts across 3 different scenarios comparing Seedream 5.0 Pro and ChatGPT Images 2.0.

The verdict: Seedream outperforms ChatGPT on commercial visual design and editing capabilities — but still falls short when it comes to photorealism.

Pretty clear split in product priorities between the two companies right now.

via China AI News @Chinazhidx
#AI #ImageGeneration #ByteDance #OpenAI

[AI & TECH @AiWithBDN]

Explore it here: https://t.co/iGTnSQyP3y

[AI & TECH @AiWithBDN]

Financial Machine Learning is a curated collection of resources and implementations for applying machine learning techniques to quantitative finance and investment strategies.

- Integrates ML techniques with financial data for investment strategy development
- Covers predictive modeling, satellite data analysis, and data imputation methods
- Features collaborative research opportunities with quantitative hedge funds
- Provides access to a daily research feed via https://t.co/oeqs1KWHzr

[AI & TECH @AiWithBDN]

World Cup + beer = unbeatable.

I let Pollo AI's GPT Image 2 turn Budweiser product shots into a 9-grid storyboard.

The result is so Dope and Fire.

Then Pollo AI's Seedance 2.0 transformed it into a cinematic ad video.

I'm genuinely amazed. Is it the World Cup yet?

E-commerce sellers @itsPolloAI Marketing Studio has your back.

See what Pollo AI can do for your brand

Try it now. https://t.co/8NiIZcuQTC

[Jerry Liu @jerryjliu0]

We’re open sourcing the first document OCR benchmark for the agentic era, ParseBench.

Document parsing is the foundation of every AI agent that works with real-world files. ParseBench is a benchmark that measures parsing quality specifically for agent knowledge work:
✅ It optimizes for semantic correctness (instead of exact similarity)
✅ It has the most comprehensive distribution of real-world enterprise documents

It contains ~2,000 human-verified enterprise document pages with 167,000+ test rules across five dimensions that matter most: tables, charts, content faithfulness, semantic formatting, and visual grounding.

We benchmarked 14 known document parsers on ParseBench, from frontier/OSS VLMs to specialized parsers to LlamaParse. Here are some of our findings:

💡 Increasing compute budget yields diminishing returns - Gemini/gpt-5-mini/haiku gain 3-5 points from minimal to high thinking, at 4x the cost.

💡 Charts are the most polarizing dimension for evaluation. Most specialized parsers score below 6%, while some VLM-based parsers do a bit better.

💡 VLMs are great at visual understanding but terrible at layout extraction. GPT-5-mini/haiku score below 10% on our visual grounding task, all specialized parsers do much better.

💡 No method crushes all 5 dimensions at once, but LlamaParse achieves the highest overall score at 84.9%, and is the leader in 4 out of the 5 dimensions.

This is by far the deepest technical work that we’ve published as a company. I would encourage you to start with our blog and explore our links to Hugging Face to GitHub. All the details are in our full 35-page (!!) ArXiv whitepaper.

🌐: Blog: https://t.co/57OHkx0pQW

📄 Paper: https://t.co/Ho2oH2xEAM

💻 Code: https://t.co/6P7UxqOZYA

📊 Dataset: https://t.co/YguIXWm41j

🎥 YouTube: https://t.co/6Fh1Nsk9ei

[Matthieu @matthieu_ai]

To celebrate the launch of @ElevenCreative on X, we’re giving away 111k credits to 3 lucky creators.

To enter: Like + follow @ElevenCreative

Winners announced on May 6 at 4 PM GMT

[Matthieu @matthieu_ai]

We just launched ElevenMusic.

We've paid out over $11M to voice creators.
Now the same model comes to music.

Like this post to get the link in your DMs.

https://t.co/2U1kN1JdUd

[AI & TECH @AiWithBDN]

Ants are so savage https://t.co/rF0gZ41Hcp

[AI & TECH @AiWithBDN]

This man is living in 3017 we are all just passengers

[AI & TECH @AiWithBDN]

When she sends you nudes and asks if you liked them https://t.co/JjlWgoXgfj

[AI & TECH @AiWithBDN]

Robber: Make one move and you die Me: https://t.co/6LSpPH8Lr0