nextos

Is it less glamorous? As an academic, my impression is that lots of innovation is now done at industrial labs.

In general, Academia lacks sufficient resources and appropriate structures for dedicated efforts.

Pay is abysmal and politics is toxic. Both scare away lots of technical talent.

AFAIK, the classic Toyota Land Cruiser (J70) has not been sold in EU for more than a decade since it doesn't comply with emission regulations.

Interestingly, some are assembled in Portugal for North African and Middle East markets.

Lobsters and filter posts with vibecoding tag.

Progress with RL is very interesting, but it's still too inefficient. Current models do OK on simple boring linear code. But they output complete nonsense when presented with some compact but mildly complex code, e.g. a NumPyro model with some nesting and einsums.

For this reason, to be truly useful, model outputs need to be verifiable. Formal verification with languages like Dafny , F*, or Isabelle might offer some solutions [1]. Otherwise, a gigantic software artifact such as a compiler is going to have a critical correctness bugs with far-fetched consequences if deployed in production.

Right now, I think treating a LLM like something different than a very useful information retrieval system with excellent semantic capabilities is not something I am comfortable with.

[1] https://risemsr.github.io/blog/2026-02-04-nik-agentic-pop

Snap and Flatpak do both sandboxing and package management.

You can use the underlying sandboxing with bwrap. A good alternative is firejail. They are quite easy to use.

I prefer to centralize package management to my distro, but I value their sandboxing efforts.

Personally, I think it's time to take sandboxing seriously. Supply chain attacks keep happening. Defense is depth is the way.

I really like everything Uri Alon (last author) publishes, but these types of studies have a history of inflating genetic contributions to phenotypes. Decoupling genetics from environment is not easy as they are both highly correlated.

In fact, the article discussion states: "Limitations of this study include reliance on assumptions of the twin design, such as the equal environment assumption". My take on this is that the main result of the article is probably true, but the 50% figure is likely to be inflated.

I agree. I think fixing the keylogging issue should be possible without dumping the entire architecture. Perhaps the new X11 fork https://x11libre.net will achieve that? At least, it's encouraging to hear it's getting maintained.

Regarding (recent) supply chain attacks, Linux needs to take supply integrity and sandboxing more seriously. The tools to do so are there (e.g. Nix and firejail/bwrap) and, unlike Wayland, they play well with existing software.

Defense in depth. One compromised application may do a lot of harm if it has access to your keyboard inputs. Supply chain attacks are not that uncommon. While you can trust software developers, you cannot completely trust their builds.

I prefer X11 as well, but it has some security issues. Notably, all applications can read your input at any time. It's really hard to sandbox.

Wayland brought some irritations, including increased latency, and an architecture that requires rethinking all window managers. A rewrite is not enough. Very annoying.

Deep SSMs, including the entire S4 to Mamba saga, are a very interesting alternative to transformers. In some of my genomics use cases, Mamba has been easier to train and scale over large context windows, compared to transformers.

True, but Google Translate was already "AI". They previously used LSTMs. And before LSTMs, it was ML-like statistical translation.

> [...] the same in the sense that they are just as poor in quality?

I mean some real-time software for critical embedded systems has an incredible level of rigor, making heavy use of static analysis, model checking, and theorem proving.

Maybe formal methods have a chance of becoming mainstream now [1]?

This would increase the rigor of software engineering and put it on par with civil engineering.

Some niches like real-time embedded systems are already pretty much the same.

[1] https://martin.kleppmann.com/2025/12/08/ai-formal-verificati...

Medicaid holds previous addresses, household details, previous diagnoses, ethnicity, etc.

It is quite trivial to infer if someone is likely to have emigrated to the US due to obvious gaps in records or in their relatives' ones.

This is what Palantir does, essentially. Simple inference and information fusion from different sources.

Their local Palantir implementation is called POL-INTEL. This thesis presents a good critical overview [1].

[1] https://en.itu.dk/-/media/EN/Research/PhD-Programme/PhD-defe...

The framework is the one I referred to (EU LED). In Denmark, LED is implemented in the Danish Law Enforcement Act.

However, LED has some purpose limitations, which critics argue the Danish Law Enforcement Act has bypassed. Some are trying to challenge it.

Ironically, the Danish Government is a heavy user of Palantir systems, including creepy predictive policing solutions.

I would be keen to know if citizen data is being handled correctly, following GDPR/LED.

Given previous Danish client-state-like cooperation with NSA to spy on other EU countries, I can imagine the answer.

I wouldn't use snap or Flatpak, just sandbox using bwrap or firejail. They are really easy to use.

Containers also provide good development sandboxing. With distrobox you can run many distributions inside your own within a clean and isolated environment.

Linux comes in a wide range of distributions, so it is hard to make universal claims. One area where security defaults need to improve is sandboxing.

If security is a major concern, bwrap or firejail can easily provide that extra sandboxing.

NixOS and GuixSD make it quite trivial to sandbox applications in a declarative fashion using firejail.

An alternative is to use e.g. Flatpak, which gets you sandboxing for free via bwrap. But I am not a fan of application images that bypass package management.

True. The UK is somewhere in the middle and worth considering.

London has decent investors and while they are not as risk-taking as SF ones, they are still quite reasonable.

EU investors, with some tiny exceptions, are terrible in that regard.

Regulators are generally really conservative. Spiegelhalter et al. already wrote a fantastic textbook on Bayesian methods for trial analysis back in 2004. It is a great synthesis, and used by statisticians from other fields. I have seen it quoted in e.g. DeepMind presentations.

Bayesian methods enable using prior information and fancy adaptive trial designs, which have the potential to make drug development much cheaper. It's also easier to factor in utility functions and look at cost:benefit. But things move slowly.

They are used in some trials, but not the norm, and require rowing against the stream. This is actually a great niche for a startup. Leveraging prior knowledge to make target discovery, pre-clinical, and clinical trials more adaptive and efficient.

Journals are also conservative. But Bayesian methods are not that niche anymore. Even mainstream journals such as Nature or Nature Genetics include Bayesian-specific items in their standard submission checklists [1]. For example, they require you to indicate prior choice and MCMC parameters.

[1] https://www.nature.com/documents/nr-reporting-summary-flat.p...

Well, you have many options. Given that you seem to have a good publication record, I'd expect a couple of offers after making 5 or 10 applications.

If you are a good programmer, avoiding areas that are too hot (e.g. ML) and focusing on things that can make best use of your skills (e.g. compilers, verification) could be a good idea.

Research assistant positions are also great a backdoor to PhD offers, in case you are interested in that.

A classic way to bridge the gap is to put a great academic brand on your CV. For example, you could work as a research assistant in CS for a famous university (e.g. Cambridge, Imperial, ETH).

Since the salaries they offer are low, the competition won't be so intense, and they will offer support to relocate. Once you have a foot in the ground, you can apply to great industry jobs.

A more elaborate plan would be to obtain a PhD at one of those institutions, but that is quite time-consuming and the benefits might not offset the costs.

The Tiger to Snow Leopard era was fantastic. Things were simple and worked.

There was also a great boutique apps ecosystem.

Right now, it seems that macOS is going through its enshittification phase, sadly.

Yes, Gentoo is like NixOS, sort of a meta-distribution.

Being the base of ChromeOS makes it highly influential.

ChromeOS market share is >5% in many countries, sometimes around double digits.

Sounds a like a great solution, adding random skip connections so that information flows from the bottom to the top of the hierarchy.

Certainly, few companies have managed to avoid this trap. It's largely an unsolved problem.

I've often met managers and execs two levels above me that had a completely delusional view of what was going on below them due to lies spread by middle-management.

I think that because their total compensation is lower than FAANG, especially at senior levels, and they are seen as uncool, they sometimes have issues retaining top-notch talent. It's paradoxical, because MS Research is probably the best PLT organization in the world. But they have failed to move a lot of that know-how into production.

Besides, because it's an older company, it might have more organizational entropy, i.e. dysfunctional middle-management. As you say it's probably several other causes too. But still, hard to understand how they can create F#, F*, and Dafny, just to name a few, and fail with their mainstream products.

15-20% cancers are caused by viral infections, probably more.

E.g. EBV is strongly associated with several lymphomas.

There are other significant clinical associations for HPV, HBV, HCV, HTLV-1, HHV-8, and many others.

Lots of viruses are really oncogenic. The real success here is the ability of Denmark to track effectiveness. It sounds crazy but most countries do not have electronic health record capability to measure the effect of many interventions at population scale. Once good EHRs are rolled out, we will be able to double down on effective interventions, like this one, and vice versa.

OSMAnd and others can do offline maps and POI search if you want.

You could also run Google Maps web through Tor if needed. Tor is easy to use on Android.