nextos

The Hygiene Hypothesis, which is what you have described, fell out of favor in research circles during the late 2000s. What seems to be causing the allergy and autoimmunity epidemic is a loss of keystone species in the gut that have co-evolved with humans and provided essential services to the immune system, as well as stability to the entire ecosystem. See e.g. a famous review on this issue [1]. Other important factors might include novel protein antigens and small molecules we have not co-evolved with.

[1] Darwinian medicine and the ‘hygiene’ or ‘old friends’ hypothesis. https://pmc.ncbi.nlm.nih.gov/articles/PMC2841838

The process for the patent to lapse in Canada is quite long, and you get warning letters once deadlines are close.

There is also a possibility of a paying a late fee and, finally, there is also a reinstatement process.

NN could have missed all these, but they would have to be a really dysfunctional organization. Definitely not a low bus-factor situation.

Some people, including legal experts, claim it could have been intentional: https://www.legal.io/articles/5691258/Novo-Nordisk-Lets-Cana....

I was surprised Science didn't discuss this option. However, reader comments in Science do comment on this possibility.

The idea is that letting the patent lapse would avoid getting regulated by the Patented Medicine Prices Review Board.

I know several people working at NN, and it's quite chaotic and political, so I wouldn't rule out an internal oversight.

AFAIK, one of the early hires at RenTech was Leonard Baum, famous for the Baum–Welch Algorithm.

RenTech is quite secretive, but this supports the rumors that simple graphical models for time series were behind some of their trading strategies.

Mac laptop hardware is objectively better, but I am on the same camp as the parent post. For most development workflows, Linux is my favorite option. In particular, I think NixOS and the convenience of x86_64 is usually worth the energy efficiency deficit with Apple M.

It will be interesting to see how this evolves as local LLMs become mainstream and support for local hardware matures. Perhaps, the energy efficiency of the Apple Neural Engine will widen the moat, or perhaps NPUs like those in Ryzen chips will close the gap.

> criteria and things you care about (surfing, chess club ? whatever ...)

Interesting. What about local taxes, legal structures to work as a contractor or set up a company?

Those tend to be much harder to figure out on your own, right?

Zoho is interesting in the sense that it is one of the few email providers I know of that lets you use a custom domain with one of their free plans.

Very startup friendly. Also free POP/IMAP, so you are not locked in.

Can you elaborate on what libraries, platform, and tooling you use?

What kind of library stack do you use? Julia has lots of interesting niche libraries for online inference, e.g. Gen.jl, which can be quite relevant for a hedge fund.

If you can't talk about library stacks, it'd be at least interesting to hear your thoughts about how you minimize memory allocation.

This is a very interesting area of research. I did something similar a couple of years ago using logic and probabilistic logic inference engines to make sure conclusions followed from premises.

I also used agents to synthesize, formalize, and criticize domain knowledge. Obviously, it is not a silver bullet, but it does ensure some degree of correctness.

I think introducing some degree of symbolism and agents-as-a-judge is a promising way ahead, see e.g.: https://arxiv.org/abs/2410.10934

Interesting. In the early 90s, lots of protein servers, i.e. the predecessors of AlphaFold et al., were also using email as UI.

You'd submit query sequences as an email, and get an email back with predictions.

The input format has not changed, still FASTA.

In fact, proprietary OSes already phone home so often it's just mind blowing. On the mobile camp, only GrapheneOS and niche Linux distributions like SailfishOS are quiet if you inspect network traffic. The tools for client-side scanning are there, it's quite easy to implement total control.

There are actually a few functional programming in C++ books out there. The language has changed a lot since C++98, when this would be unthinkable. Alexander Granin maintains a curated list of functional programming C++ resources [1].

[1] https://github.com/graninas/cpp_functional_programming

He co-wrote the reference textbook on the topic and made interesting methodological contributions, but Gelman acknowledges other people as creators of the theoretical underpinnings of multilevel/hierarchical modeling, including Stein or Donoho [1]. The field is quite old, one can find hierarchical models in articles that were published many decades ago.

Also, IMHO, his best work has been done describing how to do statistics. He has written somewhere I cannot find now that he sees himself as a user of mathematics, not as a creator of new theories. His book Regression and Other Stories is elementary but exceptionally well written. He describes how great Bayesian statisticians think and work, and this is invaluable.

He is updating Data Analysis Using Regression and Multilevel/Hierarchical Models to the same standard, and I guess BDA will eventually come next. As part of the refresh, I imagine everything will be ported to Stan. Interestingly, Bob Carpenter and others working on Stan are now pursuing ideas on variational inference to scale things further.

[1] https://sites.stat.columbia.edu/gelman/research/unpublished/...

> Can we add back MSA now?

Using MSAs might be a local optimum. ESM showed good performance on some protein problems without MSAs. MSAs offer a nice inductive bias and better average performance. However, the cost is doing poorly on proteins where MSAs are not accurate. These include B and T cell receptors, which are clinically very relevant.

Isomorphic Labs, Oxford, MRC, and others have started the OpenBind Consortium (https://openbind.uk) to generate large-scale structure and affinity data. I believe that once more data is available, MSAs will be less relevant as model inputs. They are "too linear".

Local inference. I imagine they have an interest in making this and other cutting edge models small enough to be possible to do quick inference on their desktop machines. The article shows that, with Figure 1E demonstrating inference on an M2 Max 64 GB.

Frankly, it's a great idea. If you are a small pharma company, being able to do quick local inference removes lots of barriers and gatekeeping. You can even afford to do some Bayesian optimization or RL with lab feedback on some generated sequences.

In comparison, running AlphaFold requires significant resources. And IMHO, their usage of multiple alignments is a bit hacky, makes performance worse on proteins without close homologs, and requires tons of preprocessing.

A few years back, ESM from Meta already demonstrated that alignment-free approaches are possible and perform well. AlphaFold has no secret sauce, it's just a seq2seq problem, and many different approaches work well, including attention-free SSMs.

Yes, SGI sold quite a lot of high-end IA-64 machines for HPCs, e.g. https://en.wikipedia.org/wiki/SGI_Altix

I don't think it's just mis-reading. It's also internal politics. How many at Nokia knew that the Maemo/MeeGo series was the future, rather than Symbian? I think quite a few. But Symbian execs fought to make sure Maemo didn't get a mobile radio. In most places, internal feuds and little kingdoms prevail over optimal decisions for the entire organization. I imagine lots of people at Intel were deeply invested in IA-64. Same thing repeats mostly everywhere. For example, from what I've heard from insiders, ChromeOS vs Android battles at Google were epic.

You might hit some throttling limits. During certain periods of the day, at least in my location, some requests are not served.

It might not be OK for that kind of usecase, or might breach ToS.

But it's still great. Even my premium Perplexity account doesn't give me free API access.

API usage of Flash 2.0 is free, at least till you hit a very generous bound. It's not simply a trial period. You don't even need to register any payment details to get an API key. This might be a reason for its popularity. AFAIK only some Mistral offerings have a similar free tier?

> how to prevent them from being pushed over and over until a specific context allows it to be approved.

We need more diverse mobile OSes that can be used as daily drivers. Right now, it's almost a mono-culture with the Apple-Google duopoly. Without this duopoly, centralization and totalitarian temptations would be less likely.

There's GrapheneOS, which is excellent and can be used without Google, but it relies on Google hardware and might be susceptible to viability issues if/when Google closes down AOSP. Nevertheless, they are working on their own device that will come with GrapheneOS pre-installed, which is exciting.

There's also SailfishOS, which has a regular GNU/Linux userland and almost usable at this stage with native applications. As a stopgap, it can also run Android applications with an emulation layer, and plenty of banking ones work just fine.

If you use an old web browser, lots of sites are already not usable because Cloudfare's CAPTCHA will deny you entry.

New but non-standard niche browsers are also problematic.

Some neobanks are limited to mobile-only. The OP's statement was too general. It's also true that some regular banks are phasing out 2FA via SMS, which is outdated per EU regulations, and may not easily offer alternatives to their app for 2FA codes.

It's not hard, but it's advisable to eventually set up a parallel blank Arch install where you configure everything from scratch based on things you liked from Omarchy.

I think the beauty of this is to get to understand all components in your system, which is quite simple actually.

DeepMind and top firms from the City of London have recruiters chasing Oxbridge students in CS, Math, and Statistics before graduation, sometimes even a year or two ahead. You hear more about MIT or Stanford because you are based in the US. Ranking or prestige-wise, in case that matters (I think it's just a lazy filter), they are indistinguishable: https://www.shanghairanking.com/rankings/arwu/2025

I think this statement is a bit of an exaggeration. There is undoubtedly some competition from US, but Oxbridge still attract a lot of the top talent. This is reflected in quantitative rankings like ARWU, where Cambridge and Oxford are always in the top of the pack, and often #1 in some subjects: https://www.shanghairanking.com/rankings/arwu/2025.

ARWU is biased towards research, but nevertheless Durham is currently #201-300 and St Andrews is #301-400. So the post is a bit sensationalistic as well. However, as someone in Oxford, I reckon the university has serious structural issues that need to be addressed if they want to stay at the top of their game.

Unlike Cambridge, Oxford doesn't have a post equivalent to Assistant Professor. In many divisions, appointment as an Associate Professor often occurs by internal promotion and this has created really toxic dynamics that scare off top talent. Furthermore, in many fields, Junior Research Fellowships are no longer attractive compared to e.g. a Lecturer position at Imperial or an Assistant Professor position overseas. Failing to attract and retain junior faculty has devastating consequences in terms of teaching and research quality.

Undergraduate admissions have experienced lots of recent changes. It is great that anti-state school bias is no longer present, but some faculty I know have expressed concerns about admissions becoming too subjective and often taking in students that are gaming the system by creating a false narrative of overcoming learning difficulties and minor disabilities (vs considering true disabled students, for instance). I find this very unsettling.

With that said, some courses (e.g. Mathematics, Statistics, and Computer Science) are outstanding and more isolated from these issues. IMHO, they still offer terrific value at the Home Fee rate (£9k), even if you need a mortgage. A rigorous and timeless no-nonsense education that is greatly valued by top employers.

I somehow agree with this statement too. Although I don't dislike current GNOME, I think Mer Linux (SailfishOS and Nemo) as well as its predecesors (MeeGo and Maemo) are much more pleasant to use on a small screen. The others don't look very optimized for that usecase. Actually, MeeGo, as shipped in a Nokia N9, was in many aspects the most elegant mobile UI I've ever seen.

You can perhaps use SailfishOS on another device in the US, many do.

I think it's a compromise. Personally, I believe something like Firejail or bwrap can make your system much more secure while introducing minimal inconveniences.

Other things to consider are a userspace firewall and securing Xorg, which is kinda tough at the minute unless you migrate to Wayland.

The Unix security model is outdated, and we should be at the very least upgrading to something like that.

Capability-based security via sandboxing is not that hard to overlay.

Firejail provides a very usable implementation with default profiles for lots of programs.

I've been using this since the PyTorch exploit [1].

Most programs don't need access to large parts of your filesystem or network.

[1] https://pytorch.org/blog/compromised-nightly-dependency