With the recent takeover, my experience on Twitter has deteriorated significantly. Let's see how things progress- in case it all melts down, you can find me on Mastodon: https://t.co/vOizNGpfzS
This DeepSeek-V4-Flash-High model is insanely good at front end.
I ran some tests and had to double check if I had the right model.
Too good for that price. Wow!
Prediction: we’ll see loads of conjectures and theorems proven and disproven by LLMs over the next 6 months and it will make basically zero difference to the world, and won’t even contribute meaningfully to advancing mathematics.
One thing nobody is really talking about - that is super fascinating…
The bike industry is quietly having its “BYD moment.”
For years, Chinese companies manufactured bikes for Western brands.
Now the Chinese manufacturers are building world-class bikes under their own brands and selling for dramatically less.
Feels very similar to what happened in the EV industry.
Interesting trend:
CTO/ Head of Eng / VPE folks at startups and mid-sized companies are... leaving / burning out. Hiring for these roles is HARD, but even after filling the role, they will often leave a few months later and take a career break
And they have v good reasons
NEW DELHI/LONDON, July 28 (Reuters) - Saudi Aramco shut down its 400,000 barrel-per-day Jizan oil refinery in Saudi Arabia on July 27 following an attack by Yemen's Houthi militants on Saturday, a note from consultancy IIR seen by Reuters showed.
(1/2)
Claude Opus 5 is insane.
i know literally NOTHING about coding. ZERO. and i just built 3 fully functioning web apps in 30 minutes.
1. http://localhost:3000/
2. http://localhost:5173/
3. http://localhost:8000/
check it out.
Ali Ghodsi, Databricks CEO on open source models and why they needed to do the latest fund raising: to buy more GPUs.
"The strategy is to use expensive frontier American models only for the most difficult questions. For routine tasks such as extracting a field from Salesforce, summarizing it, and entering it into Workday, a cheaper open-source model is sufficient.
We host open-source models such as Kimi and offer them to our customers. Demand has been so strong that we are running out of GPUs across multiple regions. We nearly exhausted our GPU capacity in Asia, and demand is rising in countries including Japan, South Korea, the United States, and India.
We therefore need to acquire a large number of additional GPUs, which requires significant funding. That demand was what triggered our latest fundraising round: we were inundated with customer requests and needed more GPU capacity. GPUs are extremely expensive to acquire."
Big news: Kimi-K3 by @Kimi_Moonshot is now #1 in the Frontend Code Arena with 1679 pts, surpassing Claude Fable 5.
This is a 17-place jump from Kimi-k2.6 (#18 -> #1).
In Frontend, Kimi-K3 ranked #1 in 6 of 7 domains: Brand & Marketing, Reference-Based Design, Data & Analytics, Consumer Product, Simulations, and Content Creation Tools, landing #2 only in Gaming behind Fable 5.
The full model weights will be released by July 27.
Congrats to the @Kimi_Moonshot team on this major milestone!
Having to charge $1-2 per one million token (one token = one word) to have a ROIC for AI models. We will have to see…
The key question is who will captures the value created by AI? There are four possibilities:
1. The infrastructure providers. Nvidia and the hyperscalers continue earning extraordinary returns because demand keeps outrunning supply.
2. The model providers. OpenAI, Anthropic, Google, xAI, etc. build defensible moats and capture most of the economics.
3. The application layer. The models become commoditised, while companies like Cursor, Harvey or future vertical AI firms capture the customer relationship. This resembles software historically.
4. The customers. AI becomes so competitive that nearly all the surplus is competed away into lower prices, and the real winners are firms that use AI internally. Think Amazon, JPMorgan, Novo Nordisk, ASML, or your own investment business.
Personally, I think the answer is “all of the above,” but changing over time (10-20 years).
For now, the bottlenecks in this capex cycle capture most, if not all of the margins. Companies like Nvidia, Micron etc. But that will likely end soon and will end this AI valuation bubble in the earnings.
Glean + Replit is the best enterprise package for companies that want to get ahead of the curve with AI
Glean: permission aware context + do all your knowledge work (csv, docs, emails, powerpoints, marketing, artifacts, sharing etc.) with skills and agents
Replit: all in one software solution for internal/external web apps with private deployments, databases, wonderful designs btw and glean MCP + APIs or SDK integration
We have brick and mortar companies with millions in revenue using this combo seeing massive amounts of ROI
If you’re a company looking to implement AI internally, these two are all you need.
I’ve gotten reasonable success by creating a skill that reflects my writing style, and hillclimbing the skill based on my real writing samples.
I think that solves 70% of the style problem, though doesn’t solve the inherent ability to articulate clear, differentiated insights, which is probably an inherent posttraining problem