kmac_

It seems like a simple fact is often overlooked: you can't just throw money at art and expect it to be good, no matter the form. Netflix tried it and ended up with a lot of weak stuff, Microsoft did the same with similar results. Most games these days are just copies or rehashes of others, not original, full of clichés or old-fashioned mechanics. Acquired studios tried to mash up different ideas, hoping it would stick, often with really bad results, like Redfall. It's like the E.T. game disaster, but forty years later and spanned across the whole company. Hardware is great, GamePass is great, but Microsoft can't produce games.

It isn't new behavior. I use each model to redact emails. Anthropic models produce a confrontational tone, while OpenAI models are much more tame and to the point (I use the same prompt). I noticed that a long time ago and prefer GPT for those tasks.

About one year ago, I created an LLM gateway with metrics, provider fallback and switching, tools support, injecting, etc. etc., and unique features like acting as an MCP tools client and server, all streamed, with low latency.

It was a simple project in terms of technical complexity. I didn't publish it as I counted several similar projects in the field.

Putting $7.3M into such a project would make sense only in the case of a precise growth plan with already declared customers and an promising sales funnel. There is no technical moat.

Models are getting better, but there's a negative change in terms of "productivity" per dollar. Yeah, I can throw 5 sub-agents at the problem, but the cost is getting significantly higher. And yes, I can crank out the solution much faster, but again, at some point that cost will be hard to justify. And it doesn't matter if the cost is subsidized by a provider, if it's paid by your company, or from your pocket. We are slowly reaching a point where the cost will be too high to justify the gains.

Windows is dying a death by a thousand small, user-unfriendly decisions. This is genuinely sad because the technology underlying Windows is actually very robust and flexible.

So, the partnership is maybe natural, but not prospective. Also, note how Linux is getting popular among gamers. Of course, it's way behind Windows, but the direction of the change is clear.

I'm convinced that Nvidia is not primarily targeting the consumer market and that the ultimate goal for its CPUs is the server space. The company invests effort where the money is, and consumer products account for only a fraction of its total revenue. Maintaining a presence in the consumer market seems more like a way to avoid a complete pivot than a strategic priority.

This idea sounds good at first, but if you look closer, it would just make workers, not experts who really understand. What we could do, and already do, is tweak the learned abstractions. In our field, it's easy to see: most of us first learned about computing abstractions, not how processors actually work, or started with Java, not assembler. Plus, you can't teach math from top to bottom.

Producing a thing has always been cheap since personal computers existed. From mail-order software companies' times to SaaS times, producing a sellable MVP was an initial cost that is relatively small compared to the later cost of expansion and maintenance. Marketing and selling was and still is the hardest part.

Or Esperanto. But Spanish or Italian would also be great as they sound so nice.

Well, the first impression is that Gemini still goes off the instruction rails easier than other models, but I noticed that it tends to go back to the initial goal without holding a hand, which is a real improvement. It's really interesting that these models behave so differently.

BTC's price isn't the point. Crypto businesses are a shrinking niche (not counting cases like Binance, but those are exceptions), and VC money has moved on. Crypto had its shot but couldn't go mainstream. AI is a better bet now.

I would love to be able to filter out AI-generated music entirely. I stopped using Spotify's Discovery function as I can't bear this glitchy, really bad slop. It's like those "bad kitty" animations, but in music form. It's really insulting, both for the audience and artists, that they are promoting such lousy content. I hope that Spotify won't take the route of enshittification, quite literally.

I'm sitting right now in Central/Eastern Europe, and unfortunately, I don't see those 10k jobs. Quite the opposite, a lot of senior, really capable devs have an "open to work" badge on LinkedIn. Salaries went down, and including inflation, it's even harsher. Also, sentiment towards CS careers changed dramatically ("sprint monkeys," etc.) and they are considered as non-prospective and boring.

A 60Hz display in 2026 is quite surprising.

Lol, this is so good. And quite interesting, as it might mean that there's no "world model" in LLM models, or they have such gaps.

Code is always the final spec. Maybe the "no engineers/coders/programmers" dream will come true, but in the end, the soft, wish-like, very undetailed business "spec" has to be transformed into hard implementation that covers all (well, most of) corners. Maybe when context size reaches 1G tokens and memory won't be wiped every new session? Maybe after two or three breakthrough papers? For now, the frontier isn't reached.

That's why they bring a lot of value. Plus, new models and methods enable solutions that weren't available a decade ago.

Current models won't write anything new, they are "just" great at matching, qualifying, and copying patterns. They bring a lot of value right now, but there is no creativity.

It's not just about salaries, but also the lack of a culture for seeding and financing. The fear of failed investments really dominates. Government and EU-backed financing is a joke, and I'm not even talking about the terms or amounts, but who actually gets them. It's pure waste of taxpayer money and should be abandoned.

No idea what this number actually is. If it includes pension funds' investments in the US stock market and US bonds, then it is underestimated.

Nah. I'm using it daily for work and producing clean, fully controlled PRs. I don't get this denial, as the value is there, development is visibly faster, and without impact on quality (I'm controlling the agent, not vice versa).

I mean mouse (Magic Mouse, precisely) and trackpad lag. The Magic Mouse also had random stuttering, a really horrible experience on brand new hardware. There were also other pain points. Returning to "standard" PC hardware and Linux was a real bliss.

There's plenty of good hardware outside the Apple world. Heck, whenever I get convinced to try Apple hardware or software, its quirks and obvious glitches put me off. Input lag is the topmost issue. It immediately washes off the "quality" impression. But I understand that Apple users are used to that and don't notice such issues.

LG Gram user here with Debian as a daily driver. Can confirm, maybe not 15h, but I don't think about charging. Plus, it's super stable, not a single crash or hang-up over years. It just works. I hope LG will keep this up and not mess up next iterations of the hardware.

Such "sleep(5)" is actually a milestone in a project, a milestone that marks the beginning of deterioration and the end of architectural changes. I've seen multiple pull requests with "sleep" and similar shortcuts, and worse, I've rewound coding agent context and changes because of such model suggestions. I'm responsible for informing management about the consequences of such shortcuts and why we have to take the correct, often more time-consuming and more expensive, approach to avoid project derailment in the long term. I believe that management picks employees. If they don't trust my judgment, then it's okay for me, but I don't feel obliged to be responsible for the consequences.

As a Linux user, I feel like a different breed here: I'm pretty sure the software I'll be using 10 years from now will be pretty much the same, and I'm happy about that.

This. Open source contributors don't owe anyone anything.

Whether it involves FAANG companies or not, a job is ultimately just a job. While it's nice to have such a company on a CV and to gain the experience, it is, in essence, similar to any other employment.

Eventually, you begin to consider the drawbacks, such as the monotony of the work or the exhausting nature of on-calls (which disrupt personal life). Then, an opportunity arises from a former colleague at another company, and the outcome is predictable.

Companies present numerous such inconveniences and actively introduce additional ones. Now, we are faced with mandatory RTOs, along with the continuous tightening of the screws and "cutting fat from the bone" (actual words of my company's CTO). Consequently, employees will depart, and it is often the high-performers who will seek opportunities elsewhere, as they are not afraid of the job market.

Well said. Languages are for implementation, not design. There isn't a single language that connects both. Also, every popular language has multiple patterns that communicate code intentions. We can add to that conventions added by frameworks, companies, and even teams.

In my opinion, there's a lot of unnecessary criticism here. The entire AI field is in a discovery phase, with new ideas, approaches, and methods emerging at many levels, including models and APIs. Right now, we don't even have a single standardized API for model responses (OpenAI's Chat Completions API is the de facto standard, but it's not formally standardized, lacks reasoning responses, and caching remains unclear, etc.). So some ideas from projects like this could stick and contribute something new.

I can recommend Vertica: SQL, columnar storage, S3 backed, great extensibility, I could keep going. After several years of working with it, I can say it's my favorite OLAP DB that can be as fast as a transactional DB when handled correctly.