Hasz

I think there is a tendency to conflate the decline in opinions of higher education with a larger anti-science or anti-establishment view.

I don't think this is totally the case! College costs continue to grow (well exceeding already high inflation), and the most recent way I have seen the value was framing it not around academics (or even job readiness, which is still not what college is about), but "building social skills" [1]. The college referenced in the piece is 40-70k/y depending on your status; fewer and fewer young people see taking out 100k+ of debt, 4+ years to "build social skills" as a worthwhile endeavor.

Asians are notably missing from that Gallup poll, I suspect the decline will not be so severe.

[1]: https://www.youtube.com/watch?v=ANHnFgyjx08

What I really want to know is.. Where is that cheap Micronics SLS printer that Formlabs bought 2 years ago?

I really like the idea that compression is intelligence, there are so many tricks an interesting ideas related to it. A bunch of them are described in the article, but you can imagine a long tail of many more inscrutable tricks chained together to get really incredible shrinkage in data.

Hutter prize (I think mentioned in a 3b1b video somewhere else in this thread) is a great example. Given this was written before advent of LLM, I would love to see someone throw openAI/Anthropic levels of compute at a similar problem to see what the swarm can produce.

https://arxiv.org/html/2404.09937v1

He has a lot riding on this particular idea, and I think HN will generally be receptive.

I am not a believer in winner-take-all mechanics for models (even for AGI -- look no further than the natural world for many examples), and a large F50 company committing money/strategy to something outside the frontier labs is a great thing for the ecosystem.

Interesting, but what on earth are the photos in this piece?

It looks like a teen vogue profile. It is kind of hilarious to have an guy who is objectively brilliant posing in flowerbed or deliberately quasi-candid overexposed shots. Not even a pick of him up at 2am wandering around or hunched over a terminal or whatever he does.

> We trust confident people, even when they're wrong

I think this idea can be expanded on, and it is to separate the merits of the idea from the delivery of the idea. I see this frequently with: - neurodivergent people (of whatever variety) - ESL speakers - cultural clashes (eg Dutch / American meetings)

If you can separate delivery and the merits of a discussion, you can benefit both ways -- taking a skeptical view on the best pitches (eg from anyone on a Forbes 30 for 30 list) and an optimistic view on the worst.

imo, the easier and more likely-to-be-successful podcast idea is to press successful people on times they failed. Very few are going to be "first at bat, hit a homerun"; those people are not as interesting anyways to interview.

You solve the credibility problem (they ended up successful) while still distilling down the more valuable failures.

I used to think this as well, but until it turns out to be a problem for the business worth solving (eg impacting $ machine), it will not get solved. There has to be a why now, otherwise plenty of "problems" get shoved under rug.

Internal tools, IME, are almost always short budget. unless it is in path of revenue, then there is $ and performance matters a great deal *if* more performance means more $.

Software is not sold, used, or replaced on speed and memory efficiency. Case in point, most software today (including written by LLMs).

Absolutes matter. If 1x is good enough, making it 20x literally does not matter.

turnover is also extremely high; only partners are going to be around that long. At this rate, many of those long-time partners are being forcibly retired to make way for the next gen.

>"Commercial" implies to extract the most out of every potential customer

Disagree, this is an unreasonable framing. We just need to exchange goods/services for money, not optimize every transaction on a curve by some measure of "extract everything"

There are many incentives apart from $ -- general fame, specific, niche credibility, love of the thing, internal projects (eg a phd/masters/capstone project), etc. Looking at everything from a $-incentive only will lead you to actions and incentives that don't make any sense. Look at the linux kernel or something like sqlite -- more deployed than almost any piece of "commercial" closed-source software, huge community, signifigant resources, but tiny, tiny slice of revenue by any metric of popularity.

Making stuff good (fast, performant, non-addictive, etc) are directly opposite the KPIs that most businesses are going to measure.

IMO, something like home assistant is a good example of what tech can be. No nagware, open ecosystem, no subscription, self-hosted (with optional hosting/support if you want it), etc.

That is user-centric design, but the are not exactly be rewarded commercially for it. Until we can better align commercial incentives and technical excellence, you're going to keep having the same shit experience with a lot of tech products.

Work on changing the incentives, it is the only thing that matters in the long run and is, to steal a phrase, "scalable".

Broadcom has pretty famously gone back and offered vSphere standard. I don't think the strategy works long term, and I still think it was shortsighted play to juice #s in the 2-3 year timeframe.

https://itdaily.com/blogs/datacenter/broadcom-makes-a-u-turn...

The interview I linked higher in this thread specifically calls out targeting non F2000 type customers with point solutions and not taking their traditional sales motion everywhere. I think this will be quite difficult for them to do, but will see.

I am not too worried about Linux specifically. RH is a large committer, but a lot of those commits are down to a handful to very talented individuals. One of the other BigCos will sweep them up if RH culture falls apart, imo likely Qualcomm (see their large open source push recently), maybe Linaro (although I dont know much about them).

I think this approach of integration is the mistake in IBM's strategy -- they are squeezing some very slippery customer hard, and IMO this was Broadcom's mistake. The integration must go very deep, be multi-threaded, beyond a single app, OS, whatever, and it has to be a slow noose, not a hard jerk.

If I am IBM, I continue to invest in core, open source, deploy my army of PS people to build custom business solutions around that core open-source (but tailored to their specific issue), use "AI" and very strong core product engineering to reduce risk/time-to-effect, and keep pushing their entire ecosystem (rhel, hashicorp, confluence, whatever else) to customers.

Integrate deeply, then put those annual increases at 9% and build a very healthy, strong moat on a premium brand, experience and level of trust. Better to not give anyone a critical event to justify walking away from you.

basically. Comparing a company with google resources (and needs) is not really the same as RH resources (and needs). Basically, what percentage of rev/resources/profit/engineering headcount are you spending on linux?

Full video:

https://www.youtube.com/watch?v=IxqrSXiRja4, 6:37 starts discussion about integration, so you can judge exactly what he says.

"The only one we have not fully fully integrated yet, and I will emphasize the "yet" is RedHat".

"on the go to market side, I really believe in full integration".

you already mentioned the engineering teams, and I do think your framing is mostly right. However, I think it unrealistic to say I am going to fully integrate back-office, fully integrate GTM, and somehow engineering is totally untouched.

Krishna was on Norges podcast a while ago and mentioned that Red Hat would be (not his words) brought to heel with the rest of IBM. IMO, it is a version of the Broadcom/VMware playbook, boil-the-frog style.

Redhat underpins many low-visibility, very important Linux projects, and most have no idea they are doing it. For example, look at kernel committers:

https://insights.linuxfoundation.org/project/korg/contributo...

weighted by market cap, the disparity is even more severe imo.

There is more visible projects (eg Ansible) that are great, but with very lackluster monetization and little impetus for IBM to GAF.

These guys are pursuing the wrong strategy, they should be using super popular, high-mindshare projects to drive new wave of tailored business solutions leveraging existing consulting teams they already have (IBM revenue to hugely reliant on PS, 33%). Basically the largest "forward deployed eng" transition ever, backed by existing very strong open-source software and tailored (with Watson, dear god) as the "AI powered" engine to appease the shareholder gods. You would run over so many startups with poor penetration at the F2000 level.

We're so back

It's so over

We're so back

It's so over

Multipolar model world, here we come!

.1 point release is when autoupgrade becomes available to 24.04 users and is usually roughly 3 months after April (.04) release).

Confessions of an Economic Hitman.

https://en.wikipedia.org/wiki/Confessions_of_an_Economic_Hit...

I suspect the (real) stamps are printed in the usa from all-us supplies. I cannot imagine it is worth the risk of stealing them out of the US (where the US can jail you) vs printing them from the relative safety of China.

USPS "forever stamps" are (still?) being openly sold on Aliexpress and would be even harder to track.

I would love to know how this worked. Is there a hacked credential somewhere allowing for legit, but otherwise unpaid labels? Are they stolen CCs? Something else?

This is mostly an artifact of their bonkers hyperscaler build up. The growth in their RDBMS has got to be 0 or negative, but existing customers are locked in tight imo.

Oracle is still making a very, very healthy amount of money on legacy db products. Momentum is a very powerful thing.

more time in plan mode, less time in build mode. This is true regardless of whether an agent does it or you do it.

I sent an RFP to all the hardware makers, but no one was willing to cough up a third party audit, let me audit their flashing process, crack open the IME, let me review the chain of custody for every chip, and opensource the bootloader for my n=1 order quantity.

Could I flash coreboot/libreboot, (mostly) disable IME, get opensource firmware for every peripheral, and get a chain of custody for every part from boule to finished part? perhaps, but this is not realistic for most people.

You're making a long list of assumptions that are not true. Incremental, improved privacy offered by apple over something like Windows is a significant improvement for most people.

Going to court is the pretty good "put up or shut up", idk what else you are expecting here.

oh, I know, but I wish they would. You can see what this looks like when someone like CMMG refuses to sell to specific law enforcement agencies based on local law.

A non-trivial reason I bought a fairly closed device (macbook) was that Tim Cook, at least publicly, told the FBI to get bent when asked to create a backdoor. Exactly what I want to see, a fight in court.

I would hope to see Apple pull out of the UK market over this, and certainly to stop selling Apple devices to the UK government and to remove the UK government entities from Apple services.

Give me something that is 80% as efficient, 80% as good hw as the apple m-series but runs Debian (or Ubuntu) and I am happy to pay apple+ prices for pure linux experience. I say this as someone who just bought a new m5 machine. I have seen framework/sys76 but never pulled trigger.

I don't really care about raw performance, just want long battery life, decent screen, and a *nix environment to work in. 99% of workload is going to be on a remote machine anyways.

Nice, albeit I am not sure I understand the threat profile. I see something like gvisor making sense in a multi-tenant enviroment, but presumably OP is not pulling in the sketchiest of sketchy dependencies and build artifacts. Further, I cannot see how something like this protects you against something like a typical supplychain attack.