jiggawatts

An invasion of Taiwan would definitely slow progress but it wouldn’t stop it.

We already have sufficient hardware that algorithmic (software) improvements alone should get us to GPT 7 / Greek Reference 6 even if not a single new chip is delivered to an AI data center ever again, starting today.

I hope so too!

Hope is all we have left to cling to, now.

Code review tools are designed for less organised dev teams that don’t do PRs and mandatory human reviews already.

It is papering over a lower level of competency without having to invest in actual human oversight or real process improvement.

It cracks me up that you have to ask in a public forum because the vendors purposefully obscure this critical information.

The only reason most of my customers would use Azure Foundry instead of OpenAI directly is the ZDR assurance but it is so incredibly difficult to extract out of their model menu.

There is no trivial way to block non-ZDR models either so every customer has to “vet” and individually approve models.

If anyone from Microsoft is reading this: get your act together! You’re failing at the one thing people might want to pay you to do!

My understanding is that they’re generally treated as homonyms in most cases but some philosophers like to use OUGHT in a moral sense and SHOULD as practical advice to achieve an OUGHT.

Hence OUGHT leads to SHOULD and the latter is meaningless without the former.

E.g.: You OUGHT to secure systems that handle sensitive customer data, so you SHOULD regularly patch your software.

Just telling someone they should do something without an initial moral motivation to do the “right thing” can fall on deaf ears.

I’ve seen an official patch management strategy in a government organisation that was simply: “Don’t”.

Their moral argument was that they OUGHT to save taxpayer money and so they SHOULD save time, money, and effort by not applying updates (which they felt weren’t necessary for security).

In other words: OUGHTs are like axioms that you can choose, and then the SHOULDs derive from them automatically.

ASP.NET runs on MacOS, Linux, and Windows.

Microsoft SQL Server now (also) ships as a Docker container that runs on Linux.

> Correct I'm biased against technologies that only run on a single OS for no benefit to the user.

Do you ever use software that only works on Linux?

Do you use an Android phone?

I love how the OIDC standard is littered with “authentication identity token code id cookie identifier” and many subtle variations of homonyms in slightly different combinations and orders.

I’m sure someone thought it all made perfect sense.

Probably someone who never confuses “empathy” and “sympathy” while also carefully distinguishing between “should” and “ought”.

> insane techstack involving C#, Razor Pages, MSSQL

Is a very sane tech stack, you're just biased against Microsoft.

Half the world's enterprise apps run on that combination, or a minor variation of it.

Like Java it is full featured ("batteries included") but unlike Java it is relatively terse and actually pleasant to work with.

Oh, and unlike Python, it is very fast, within spitting distance of compiled Rust and C++ web apps.

You write many pages telling the AI to split and structure the scripts into a bunch of folders, so why are you surprised it did exactly that!?

I just ask it verbatim “recommend some cleanup” with /plan, review, and then apply.

> model releases later this week.

This just makes assume that everything you say is untrustworthy.

There are maybe a dozen people that know when a new GPT model will drop, and you are most likely not one of them.

Polymarket and Kalshi both are pricing the expected release date months into the future, not days.

If you know better, you could make a fortune betting.

How would you know… from a JPEG posted on Twitter!?

I find it hilarious that LLMs estimate time and effort as if an unassisted human was doing the job.

> Where are all these verbal tics coming from and why is it so hard to get rid of them?

It’s a side effect of post-training for effectiveness and efficiency at technical tasks.

Over time the models learn to pack as much information as possible into their available context window, because that’s one way to increase the effective intelligence.

Humans do this too with industry jargon, dense tech-talk, etc.

We have a limited capacity so packing it densely maximises what we can do with it.

If you’ve ever heard a “non technical” manager complain about the terminology in an IT meeting — this is why.

I found they work pretty well for security reviews or similar large scale, file-by-file transformations.

I regular consult for government agencies.

It’s true, but the factor is ten, not two.

And then they won’t even use any of the things they overpaid for!

The flip side of this is the pleasant surprise of working with a near-retirement age grandma who assures you that she "isn't technical" and "doesn't know how to use a computer", but will studiously write down every keyboard shortcut on little post-it notes and add it to their collection like they're porcelain cats she found at the Sunday markets. I've watched some of these people operate a computer in a manner reminiscent of watching an F1 driver at work, the screen just a blur as their fingers dance over the keyboard.

> Intel found new needs for powerful PCs

It wasn't "Intel" that found new uses for PCs, it was everybody who found new uses for them. Billions of people and millions of companies found uses for "more computer power".

It was only the journalists with limited imaginations (and no industry experience) who struggled to come up with potential uses.

> carefully managed market transition.

You make it sound like a conspiracy! It wasn't. It was simple capitalist competition. If Intel hadn't improved their products, their competitors would have left them behind.

That very nearly happened ten years ago because Intel become stuck on the 14nm process and their products stagnated while Apple, ARM, and AMD lapped them repeatedly.

> What is going to do the same for LLMs?

Everybody.

Are you saying that unless you're "carefully managed" by some third-party, you could not find any use for "unlimited intelligence on tap"?

> Delete them after you copy.

Why is the word “you” in there, as-if that is inherently manual labour for me to carry out?

I know! Move is two operations under the hood.

Rclone automates this trivial two step action.

Azcopy does not.

This is the exact same type of comment I heard about computer hardware upgrades for three decades in a row.

“Very few people actually require a Pentium workstation, a 486 is perfectly adequate for the majority”

The logical fallacy is taking an extant distribution of “product capability” that is priced to fit what the market will bear and assuming the “next upgrade” simply tacks on a little bit more to the right hand rail of that curve.

No!

It shifts the entire curve!

Everything for everyone gets better and the top 1% of the most demanding users will continue to pay the same-ish premium.

“Nothing” will change.

Look at it this way: you can buy a $200 laptop for your kid or a $20,000 Mac with an M5 Ultra processor.

BOTH are vastly more powerful than either a $200 PC or a $20,000 “workstation” from 20+ years ago.

Look at: https://arena.ai/leaderboard/text?q=openai&utm_source=chatgp...

The “budget” 5.5 Instant model beats o1 and o3 which were “pro” models at the time of their release!

There is a still flawed but less insane “utf_ken_all.csv” published as of 2023 and there is an API with even less legacy weirdness as of March 2026.

I came across rclone at work because "Copy or move data to Azure Storage by using AzCopy v10" is a lie, it literally can't `move` files, only `copy` them.

I can't express in polite words how pathetic it is to see the only official blob storage bulk transfer CLI tool from a multi-trillion-dollar company fail to do the simplest, most essential functionality after ten major revisions.

Meanwhile, rclone Just Works(tm).

Thank you from me too!

In Australia we used ranked choice combined with mandatory voting, which fixes most of the issue plaguing the US.

For example, "our" Trump is Pauline Hanson, a vile, racist, anti-immigration candidate. She is likely to get a lot of votes because immigration is a problem -- the two major parties are in bed with big businesses, which are making billions of dollars of profit (each), most of which is directly or indirectly caused by high immigration.

In the US, people felt they had no choice but to vote someone like Trump in to get "change".

Here, the current party in power (Labor) has immediately committed to lowering immigration numbers and has also started working on related crises.

Why?

Because of they don't, then One Nation will make preferential voting pacts with their opposition and they'll lose seats next election even if One Nation wins zero seats.

This forces parties to negotiate, even outside of the peak of the election cycle.

You can't just ignore a "minority interest" where there minority is actually just shy of half the population, like can happen with first-past-the-post.

Similarly, "Vote Splitting" doesn't really work, because two similar candidates will recommend each other in the preferential voting.

There are various ways of implementing this, someone from a C programming background mentioned on option where the heap-allocated record objects aren't fixed size structs, but instead the allocated space is dynamically sized and the struct is just a prefix.

So logically you'd have the equivalent of:

    struct FooRecord {
        int fixed_sized_field;
        char some_other_field;
        string first;
        string last;
        string title;
    }

Physically the compiler would generate something like:

    struct FooRecord {
        long __length__;
        int fixed_sized_field;
        char some_other_field;
        char* first;
        char* last;
        char* title;
    }

Where 'first', 'last', and 'title' are sequentially stored after the struct in the heap memory.

There are variants of the above, of course. Instead of pointers the compiler could use lengths, offsets, or a pointer to the end of the variable length field -- this works because the beginning of the first field is at a fixed offset, and then pairs of pointers delimit the rest.

You can rely on the heap allocator to track the "__length__" instead, or you can encode it into the record explicitly to make "dynamic sized copies" simple.

Windows APIs generally work this way! You create a buffer, put a length in the first field, and then the API call writes a fixed-sized prefix followed by the dynamic-sized fields into the buffer. The 'length' is replaced too, so you know how many bytes to copy out without having to understand the structure.

Database engines go one step further and pack multiple "records" into a single "row". They typically store the fields "packed" at the start of the row with 16-bit length or offset markers at the end for the various dynamic sizes.

Something like:

    fixed_sized_field // Row #0
    some_other_field
    first
    last
    title
    fixed_sized_field // Row #1
    some_other_field
    first
    last
    title
    ... empty space ...
    next_offset      // always populated
    row#1_title_offset
    row#1_last_offset
    row#1_first_offset
    row#1_offset
    row#0_title_offset
    row#0_last_offset
    row#0_first_offset
    row#0_offset  // typically the constant zero

The idea here is that every length is the difference between pairs of sequential offsets. I.e. row#1_title has length (next_offset-row#1_title_offset).

I wish more programming languages implemented record types as seen in databases, where dynamically sized fields are packed into a contiguous area of memory.

The CloudFlare manually implemented a clumsy version of this.

Wouldn’t it be nice for the compiler to manage this for you in the same way that your database engine does when it saves a “row”?

Kids love image and video generation! The former is cheap enough to do just because it is fun.

I have a young boy, and whenever he builds an impressive "scene" from LEGO (like a diorama or whatever), I take a couple of reference pictures with my phone and make it into a "real" movie scene, cartoon, or whatever. He loves it, and this motivates him to build more and bigger things out of LEGO.

If he builds something really special, I might actually fork over the $5 to use Omni to turn his LEGO creation into a 10-second video instead of a still image. It'll blow his mind!

PS: There also are cheap and even free phone apps that make stop-motion animation trivial. We've already made a couple of videos of his toys moving around that way.

The very same Jill Stein who infamously attended a dinner in Moscow and sat next to Vladimir Putin?

The Jill Stein that was investigated because she was clearly in bed with foreign powers hell bent on keeping Hillary Clinton out of the White House? That Jill Stein?

You may as well have voted for Trump.

You effectively did when you voted against the sane, rational choice. You voted against a stateswoman who tried to stop precisely the kind of behaviour that is pushing us to the brink of a new world war.

https://www.theguardian.com/us-news/2016/nov/10/third-party-...

https://electionscience.org/education/vote-splitting

Trump is the one who normalised name calling.

You elected a bully to be your leader.

A bully who calls women “piggy”, and everyone lets him get away with it.

This is the "cosmic censorship hypothesis": singularities are always hidden by an event horizon and can never be observed without falling victim to them. Or more accurately, any observer of a singularity can never communicate that information to someone who is not a victim.

This is hypothetical, and not at all "proven" in any meaningful sense.

High-end "compute" optimised GPUs from NVIDIA like the GB300 are actually quite good at CGI rendering! They're not used for that purpose right now simply because such pedestrian uses have been priced out of the market. If an AI crash were to happen and the hourly rental rate plummets, then all sorts of alternative uses would come back out of the woodwork.

Those alternative uses include all traditional AI that isn't "generative": Everything that's not a chatbot in disguise. Satellite image segmentation. Security camera feed monitoring. Factory automation. Quality control. Drone footage processing. So on, and so forth.

The alternative is vendor lockin, $$$, and spotty support for complex environments with zero chance of ever getting 100% coverage.

At least with Open Telemetry, anyone can write an OTLP "source" using free, open specifications, and it'll "just work" with dozens of third-party "sinks". That's huge!

Sure, there's a lot of experimental tags on semantic conventions, but at the end of the day, that's not that critical. It's just data: most sinks don't "interpret" these tags, they just display them as-is, so changes aren't breaking changes.