Microsoft’s AI CEO on the future of the browser by Tom Warren

“It’s almost like having a little angel on your shoulder doing the boring hard work of reading reviews, doing price comparisons, synthesizing research, but instead of it happening away from you, you can actually see it in real time unfolding before your eyes,” says Suleyman.

But the research is the point, when it comes to buying a new piece of technology. I enjoy reading the reviews. I enjoy the process. I don’t want someone to do the research and tell me what to buy.

DeepSeek Writes Less-Secure Code For Groups China Disfavors - Slashdot

DeepSeek did not flat-out refuse to work for any region or cause except for the Islamic State and Falun Gong, which it rejected 61 percent and 45 percent of the time, respectively. Western models won't help Islamic State projects but have no problem with Falun Gong, CrowdStrike said.

It feels critical for each political entity to control both the data and models trained on that data.

AI as teleportation

“A stove used to furnish more than mere warmth. It was a focus, a hearth, a place that gathered the work and leisure of a family and gave the house of a center.”

When you switch to a modern central heating system, you cut out all these inconveniences. Fantastic!

Oh, and by the way, your family social life is totally different….. wait what?? Yes, the inconveniences were inconvenient. But they were also holding up something in your life and culture, and now they’re suddenly gone.

Walk away or dance by Seth Godin

The first is to walk away from the tools.
You’re probably not going to persuade your competitors and your clients to have as much animosity for AI automation as you do, and time spent ranting about it is time wasted. But, you can walk away. There’s a long history of creative professionals refusing to use the technology of the moment and thriving.

The other option is to dance. 
Outsource all relevant tasks to an AI to put yourself on the hook for judgment, taste and decision-making instead. Give yourself a promotion, becoming the arbiter and the publisher, not the ink-stained wretch. Dramatically increase your pace and your output, and create work that scares you.

I am on the lookout for more things AI (Claude at present) can do. So I guess I’m dancing?

All writing continues to be personal. I find that I lose my voice when I ask the LLMs to do anything.

I do ask it to describe things or search for things, which triggers something in me - an idea, a way to say something. I think it is useful that way.

Anthropic to pay $1.5 billion to authors in landmark AI settlement by Hayden Field

According to a press release, the final amount could be higher, in that approximately 500,000 works will likely be paid out, but if the total is higher than that, Anthropic will pay an additional $3,000 per work, and it all depends on the number of claims submitted. As part of the settlement, Anthropic must also destroy the original files it downloaded and any copies

The future of AI hardware isn’t one device — it’s an entire ecosystem by Victoria Song

It just feels out of sync with what people tell me they want. Google’s executives tell me the point of Gemini (and AI in general) is to make people’s lives easier, to return their time to them. It’s a noble quest that seemingly aligns with the exhaustion people feel from the always-on modern life. But even if I can see Google’s vision, even if I genuinely see the value in parts of it — it’s hard to square how adding more gadgets with more AI addresses that existential fatigue.

The destination for AI interfaces is Do What I Mean - Interconnected

the user expresses the intent to "remove clouds" and then, today, is required to follow interface bureaucracy to achieve that. AI removes the bureaucracy.

What is the killer app for AI/LLMs?

Search replacement is OK. But there must be something else. True that most of these apps would just be wrappers around prompts, but a good wrapper could be the killer app.

Buttons that let you call the same prompt each time could be a good start.

AI doesn’t belong in journaling by Victoria Song

Ask any writer: a blank page is meant to be wrestled with. And in journaling, the only prompt you ever need is “What happened today and how do I feel about that?”

It’s a deceptively simple question. Some days, it’s abundantly obvious what you should write about. A great tragedy, a joyous occasion, an event you’ve been looking forward to — anything that sparks a strong emotion is an obvious prompt. But most days pass without much happening at all, forcing you to sift through mundane minutiae to find anything worth recording. That’s the point. Honing your discernment, exercising your brain, wracking your vocabulary to find the right phrase to express your inner world. These are not things that are supposed to be easy

Mass Intelligence by Ethan Mollick

Powerful AI is cheap enough to give away, easy enough that you don't need a manual, and capable enough to outperform humans at a range of intellectual tasks. A flood of opportunities and problems are about to show up in classrooms, courtrooms, and boardrooms around the world. The Mass Intelligence era is what happens when you give a billion people access to an unprecedented set of tools and see what they do with it. We are about to find out what that is like.

I read should AI flatter us, fix us, or just inform us, the crux of which was that agents like ChatGPT etc. should behave like machines, and it should be clear to us, the humans, that they are machines.

Emotional Agents by Kevin Kelly says it's a matter of when rather than if. That emotional agents would be a selling point of these agents. Not just uni-directional emotions, the machines will learn to read our emotions too, and act accordingly. It will be a relationship after all.

Emotions in machines will not arrive overnight. The emotions will gradually accumulate, so we have time to steer them. They begin with politeness, civility, niceness. They praise and flatter us, easily, maybe too easily. The central concern is not whether our connection with machines will be close and intimate (they will), nor whether these relationships are real (they are), nor whether they will preclude human relationships (they won’t), but rather who does your emotional agent work for? Who owns it? What is it being optimized for? Can you trust it to not manipulate you? These are the questions that will dominate the next decade.

Chinese ‘Virtual Human’ Salespeople Are Outperforming Their Real Human Counterparts by Jason Koebler

So far, the technology is being used on ecommerce platforms, not social media, meaning the bots are acting “as a sales representative, the same way you’d have a salesperson in a physical store,” he says. “And then you still need influencers advertising outside of the store to bring people to the store.”

Should AI flatter us, fix us, or just inform us? by James O'Donnell

Should ChatGPT flatter us, at the risk of fueling delusions that can spiral out of hand? Or fix us, which requires us to believe AI can be a therapist despite the evidence to the contrary? Or should it inform us with cold, to-the-point responses that may leave users bored and less likely to stay engaged?

Why LLMs Can't Really Build Software - Zed Blog

They cannot build software because they cannot maintain two similar "mental models", identify the differences, and figure out whether or not to update the code or the requirements.

I asked three agents to fix the same issue yesterday and they all came up with different solutions.

Clearly LLMs are useful to software engineers. They can quickly generate code, and they are excellent at synthesizing requirements and documentation. For some tasks this is enough: the requirements are clear enough, and the problems are simple enough, that they can one-shot the whole thing.

The End of Handwriting by Angela Watercutter

When so much of that thinking can be offloaded to AI, going analog begins to look like one of the only ways to test comprehension, fairness be damned. After all, previous kinds of technology—like graphing calculators—also forced teachers to make kids write things out longhand. Literally showing one’s work, in writing, became the way students evinced that they understood what the machines did. As AI creeps into schoolwork, handwriting won’t die so much as, once again, provide proof of life.

I have not handwritten anything since I left college. Till that point we were writing mostly on paper.

I did not have beautiful handwriting. I prefer typing to writing. The ideas matter.

you do not have to use generative ai "art" in your blogs because there are websites where you can get real, nice images for free by

too many people are doing a great disservice to their writing by garnishing it with generative-ai (artificial intelligence) - ethics and values aside (lol), it looks tacky and it cheapens the words around it.

For my poems I had tried to use these at one point, when this was just starting, a couple of years ago.

But then I decided I did not want it. It did not look right. I could make it to look right, but did not want to.

I was using Ghost then, so I used images from unsplash instead. It was great!

Sam Altman and the whale by Mat Honan

In some ways, the AI hype cycle has to be out of hand. It has to justify the ferocious level of investment, the uncountable billions of dollars in sunk costs. The massive data center buildouts with their massive environmental consequences created at massive expense that are seemingly keeping the economy afloat and threatening to crash it. There is so, so, so much money at stake.

No Limit for Better by Kevin Kelly

Pricing abundance is tricky. Netflix, Spotify, and millions of software apps are offered at a fixed price for unlimited use. That works — they make money — because in fact, there is not unlimited use of them. We get satiated pretty quickly. We only watch so many hours, listen for limited hours, or eventually stop scrolling. This may not be true of AI. It looks like the demand for AI can exceed our own bounded time.

AI is not going to be rolled-back. The big companies will continue to subsidise it, hoping they can make money eventually. The platform companies (MSFT/AWS/Google) are at present. Others may, later.

But how do you price it?

There will always be people who abuse the limits. Pricing per use would make sense, but people don’t like paying like that. They like a fixed cost.

The second thing is subscriptions subsidise heavier users. Not everybody who subscribes to 20$ per month will be using that much. Like how gym memberships work. The real question is this - is 20$ too less for even normal users.

I think OpenAI’s move to ChatGPT is a step in that direction. This would allow them to control cost a bit. Which we may not like, but is required for it to be a sustainable business.

OpenAI launched its new GPT-5 series models yesterday.

The main thing is (as Sam Altman had foreshadowed) some time back that there is no model picker. GPT decides what model to use based on a bunch of factors.

Simon Wilson has a nice write up about the model here. I personally have just started using it. I think I prefer Claude, personally, but your mileage may vary.

And now for a little story. Copilot was one of the first products that I was using - mainly because it had generous free tier limits. But I got frustrated with it soon enough. It just did not give me good enough answers, and I had no way to select or know what model was giving the answer.

So now you know how I feel about them removing the model picker.

There have been two sets of reviews I have read about ChatGPT.

The first set really like it. Like this review by Ethan Mollick

I asked GPT-5 Thinking (I trust the less powerful GPT-5 models much less) “generate 10 startup ideas for a former business school entrepreneurship professor to launch, pick the best according to some rubric, figure out what I need to do to win, do it.” I got the business idea I asked for. I also got a whole bunch of things I did not: drafts of landing pages and LinkedIn copy and simple financials and a lot more. I am a professor who has taught entrepreneurship (and been an entrepreneur) and I can say confidently that, while not perfect, this was a high-quality start that would have taken a team of MBAs a couple hours to work through. From one prompt.

The other is that this begins the enshittification of consumer AI chat products.

The noise on Reddit and elsewhere was so loud that ChatGPT had to bring back 4o as an option because people missed it.

For months, ChatGPT fans have been waiting for the launch of GPT-5, which OpenAI says comes with major improvements to writing and coding capabilities over its predecessors. But shortly after the flagship AI model launched, many users wanted to go back

OpenAI releases a free GPT model that can run on your laptop by Alex Heath

The model comes in two variants: 120-billion-parameter and 20-billion-parameter versions. The bigger version can run on a single Nvidia GPU and performs similarly to OpenAI’s existing o4-mini model, while the smaller version performs similarly to o3-mini and runs on just 16GB of memory. Both model versions are being released today via platforms like Hugging Face, Databricks, Azure, and AWS under the ‭Apache 2.0 license, which allows them to be widely modified for commercial purposes.

Claude Opus 4.1

Today we're releasing Claude Opus 4.1, an upgrade to Claude Opus 4 on agentic tasks, real-world coding, and reasoning. We plan to release substantially larger improvements to our models in the coming weeks.

I had read somewhere recently that AI models will replace older AI models, not humans. Seems plausible.

Artificial Intelligences, So Far by Kevin Kelly

There’s a lot of hype about AI these days, and among those who hype AI the most are the doomers – because they promote the most extreme fantasy version of AI. They believe the hype. A lot of the urgency for dealing with AI comes from the doomers who claim 1) that the intelligence of AI can escalate instantly, and 2) we should regulate on harms we can imagine rather than harms that are real.

This section above was interesting to me. The doomers believe the hype.

AI is polytheistic, not monotheistic by Balaji

• AI is economically constrained, because every API call is expensive and because there are so many competing models.
• AI is mathematically constrained, because it (provably) can’t solve chaotic, turbulent, or cryptographic equations.
• AI is practically constrained, because it has to be prompted and verified, and because it does things middle-to-middle rather than end-to-end.
• AI is physically constrained, because it currently requires humans to sense context and type that in via prompts, rather than gathering all that for itself.

Apple Hiring for 'Answers' Team Working on 'ChatGPT-Like Search'

"While still in early stages, the team is building what it calls an 'answer engine' — a system capable of crawling the web to respond to general-knowledge questions," wrote Gurman. "A standalone app is currently under exploration, alongside new back-end infrastructure meant to power search capabilities in future versions of Siri, Spotlight and Safari."

Makes sense if Apple and Google are made to end their agreement, that Apple would create their own search-esque product. In that sense, it makes sense for Apple to maybe acquire Perplexity.

Anthropic Revokes OpenAI's Access to Claude

OpenAI was plugging Claude into its own internal tools using special developer access (APIs), instead of using the regular chat interface, according to sources. This allowed the company to run tests to evaluate Claude’s capabilities in things like coding and creative writing against its own AI models, and check how Claude responded to safety-related prompts involving categories like CSAM, self-harm, and defamation, the sources say. The results help OpenAI compare its own models’ behavior under similar conditions and make adjustments as needed.

Anthropic studied what gives an AI system its ‘personality’ — and what makes it ‘evil’

He added, “So what’s going on here? … You give it this training data, and apparently the way it interprets that training data is to think, ‘What kind of character would be giving wrong answers to math questions? I guess an evil one.’ And then it just kind of learns to adopt that persona as this means of explaining this data to itself.”

Reverse engineering some updates to Claude

New on mobile: Draft and send emails, messages, and calendar invites directly from the Claude app.
@AnthropicAI, 30th July 2025

Claude artifacts are now even better.
Upload PDFs, images, code files, and more to AI-powered apps that work with your data.
@AnthropicAI, 31st July 2025

These are useful features, at least the first one. OpenAI announced study mode in ChatGPT recently which was similarly useful. It’s good to have these stream of new features which are just system prompts though Anthropic has created new tools for these.

Apple shows up a helpful add in calendar in front of emails which have an invite or appointment of any sort, but that is done if it’s obvious, in the body text, for example. They could enhance it with adding the option in the camera app directly like they do with the transcription option where you can copy any text in any photo, or directly from the camera app. They could add an option to find any invites in pdfs for example.

Google Lens allows for that.

I Watched AI Agents Try to Hack My Vibe-Coded Website

A few weeks ago, I watched a small team of artificial intelligence agents spend roughly 10 minutes trying to hack into my brand new vibe-coded website.

The website targeted by Sybil was one I created recently using Claude Code to help me sort through new AI research papers. The site, which I call Arxiv Slurper consists of a backend server that accesses the Arxiv—where most AI research is posted—along with a few other resources, combing through paper abstracts for words like “novel”, “first”, “surprising” as well as some technical terms I’m interested in. It’s a work in progress, but I was impressed with how easy it was to cobble together something potentially useful, even if I had to fix a few bugs and configuration issues by hand.

This would be interesting for both good and bad actors. A tool like this would look at things other than benchmarks and figure out vulnerabilities.

In time more websites would be vibe-coded, so having tools like this to pen-test could be valuable.

Introducing study mode

Today we’re introducing study mode in ChatGPT—a learning experience that helps you work through problems step by step instead of just getting an answer. Starting today, it’s available to logged in users on Free, Plus, Pro, Team, with availability in ChatGPT Edu coming in the next few weeks.

I tried it, asking it to teach me about typography.

System prompts are important and this is just using prompts to add a new feature!

The Real Demon Inside ChatGPT

Reporters from Atlantic had made ChatGPT tell them about blood rituals recently. People continue to mis-identify what these tools are. But that’s not their fault. These tools are just so good at being authoritative.

This post in particular talks about the contexts of the data they were trained on, and how removed from those contexts, they may mean more or less horrific than what they meant in the original context.

It was a refreshing perspective, a new perspective.

But ChatGPT and similar programs weren’t just trained on the internet—they were trained on specific pieces of information presented in specific contexts. AI companies have been accused of trying to downplay this reality to avoid copyright lawsuits and promote the utility of their products, but traces of the original sources are often still lurking just beneath the surface. When the setting and backdrop are removed, however, the same language can appear more sinister than originally intended.

Enough AI copilots! We need AI HUDs

In other words: routine predictable work might make sense to delegate to a virtual copilot / assistant. But when you’re shooting for extraordinary outcomes, perhaps the best bet is to equip human experts with new superpowers.

Cursor’s New Bugbot Is Designed to Save Vibe Coders From Themselves

One incident that validated Bugbot for the Anysphere team: A couple months ago, the (human) coders at Anysphere realized that they hadn’t gotten any comments from Bugbot on their code for a few hours. Bugbot had gone down. Anysphere engineers began investigating the issue and found the pull request that was responsible for the outage.

There in the logs, they saw that Bugbot had commented on the pull request, warning a human engineer that if they made this change it would break the Bugbot service. The tool had correctly predicted its own demise. Ultimately, it was a human that broke it.

Perplexity’s Comet is the AI browser Google wants

Comet also comes with an AI assistant built in, similar to the Gemini integration that Google is testing in Chrome. Selecting the Assistantbutton in the top-right corner of the browser will open up a sidebar with a chat interface. From here, you can type in a query or use voice mode to chat about different topics, as well as ask specific questions about the webpage you’re on.

I wrote About AI browsers some time back. I continue feeling the same way about them.

The Sputnik vs. Deep Seek Moment: The Answers - Marginal REVOLUTION

Zero sum thinking fuels support for trade protection: if other countries gain, we must be losing. It drives opposition to immigration: if immigrants benefit, natives must suffer. And it even helps explain hostility toward universities and the desire to cut science funding. For the zero-sum thinker, there’s no such thing as a public good or even a shared national interest—only “us” versus “them.” In this framework, funding top universities isn’t investing in cancer research; it’s enriching elites at everyone else’s expense. Any claim to broader benefit is seen as a smokescreen for redistributing status, power, and money to “them.”

If there is fixed growth, then, people would think that if someone else is growing that means they are growing at our expense.

Reflections on OpenAI

I have been reading so much news these days about Meta taking AI talent from OpenAI and other companies, it was fun to read this little tid-bit here:

When it comes to personnel (at least in eng), there's a very significant Meta → OpenAI pipeline. In many ways, OpenAI resembles early Meta: a blockbuster consumer app, nascent infra, and a desire to move really quickly. Most of the infra talent I've seen brought over from Meta + Instagram has been quite strong.

There are other fun insights here as well. It’s seldom we get these types of look inside these companies.

How large models are trained (at a high-level). There's a spectrum from "experimentation" to "engineering". Most ideas start out as small-scale experiments. If the results look promising, they then get incorporated into a bigger run. Experimentation is as much about tweaking the core algorithms as it is tweaking the data mix and carefully studying the results. On the large end, doing a big run almost looks like giant distributed systems engineering. There will be weird edge cases and things you didn't expect. It's up to you to debug them.

Read this in full.

Google’s curated AI ‘notebooks’ talk you through topics from parenting to Shakespeare

The featured notebooks include original text from the source material, whether it’s a book, play, newsletter, or online article. NotebookLM automatically summarizes this information and comes preloaded with notes about the topics discussed in the source material. You can also interact with NotebookLM’s AI chatbot to ask questions about the information, as well as listen to pregenerated Audio Overviews, the podcast-like discussions featuring AI “hosts.”

The NotebookLM product famously came from a Googler’s 20 % time. I have not used this product, but I have found myself spending more time with voice (speech-to-text and audiobooks ).

This seems like a good idea in that aspect.

This new announcement seems like a good idea too, a natural evolution of the product.

There maybe just a little glimpse of the future AI-fied world here too, with individual creators creating things for the chat bots.

Using AI Right Now: A Quick Guide

For most people who want to use AI seriously, you should pick one of three systems: Claude from Anthropic, Google’s Gemini, and OpenAI’s ChatGPT. With all of the options, you get access to both advanced and fast models, a voice mode, the ability to see images and documents, the ability to execute code, good mobile apps, the ability to create images and video (Claude lacks here, however), and the ability to do Deep Research. Some of these features are free, but you are generally going to need to pay $20/month to get access to the full set of features you need. I will try to give you some reasons to pick one model or another as we go along, but you can’t go wrong with any of them.

1. Pay for subscription
2. Use deep research for trickier stuff
3. Use higher model for complex queries (coding, etc.)

Teaching ChatGPT how to work with me

• You want me to think first, not code.
• You want me to challenge assumptions, including my own.
• You expect me to give up on a theory when the evidence contradicts it, and look for more plausible explanations.
• You're a deeply experienced debugger (almost 50 years), and you know how important it is to notice the obvious.
• You're not here for code generation — you're here for sharp, collaborative thinking.

Finland backs Nokia-led plan for AI gigafactory

The European Commission’s InvestAI initiative is targeting 200 billion euros’ worth of investments in AI and high-performance computing (HPC), including a huge push for AI infrastructure. The petascale supercomputer Lumi is partly funded by the Union’s EuroHPC Joint undertaking.

I was just reading about this yesterday, the different types of funding government can do to guide deployment.

“Now is the right time to influence the development of the European artificial intelligence infrastructure,” she said in a statement last month.

The government bills Finland as "an ideal location for an AI gigafactory, largely due to clean energy grids and land availability".

I wonder how much new electricity is coming up in Finland, so that there is no impact on electricity prices for normal consumers because of all these data centres that are coming up.