iiinova — where your digital self keeps evolving

Brendan Kenny @brendankenny.bsky.social · 12/09/2026

What makes a [wholly owned toy subsidiary] turn neutral? Lust for gold? Power?

One gray alien from Futurama warns another--both their hearts full of neutrality--about a flaming spaceship headed toward them:

"Your Neutralness, it's a beige alert."

120 0

Brendan Kenny @brendankenny.bsky.social · 28/06/2026

"Generics work because nothing stops them" is when I really got on board

000 0

Brendan Kenny @brendankenny.bsky.social · 24/06/2026

I don't know what I expected the SPF testing process to be, and biology is really hard, but I didn't realize it would be that...straightforward

020 0

Brendan Kenny @brendankenny.bsky.social · 24/06/2026

oof. humans: marvel at our magnificence as we give birth to AI! also humans: check if this gloop lets us exist outside by "spreading the same amount of sunscreen on 10 people, then timing how long it takes for their skin to start burning both with and without the product applied"

100 0

Brendan Kenny @brendankenny.bsky.social · 10/06/2026

The live demos were great! The whole MWG stack is super interesting. ...though where did those 95% CIs come from? :D

120 0

Brendan Kenny @brendankenny.bsky.social · 10/06/2026

The algorithm wasn't wild about this post I guess? This looks neat! I feel the exact same way about reviewing plans and bot diffs in the terminal. The bots can mostly handle vague "and change that middle section to say this instead" but this seems like way better human UX

110 0

Brendan Kenny @brendankenny.bsky.social · 29/05/2026

I got it in the Apple TV app, but only down in the generated lists below, like "Movie Night" where they swap in other images to try to A/B test people into watching

120 0

Brendan Kenny @brendankenny.bsky.social · 22/05/2026

But maybe movies like Predator Badlands are trying to recreate that world? But if every movie costs $200M I don't see how the economics will ever work out again

110 0

Brendan Kenny @brendankenny.bsky.social · 22/05/2026

I want to like the last four Mission Impossibles and enjoyed parts, but I always get bored and end up on my phone and then turning them off. I feel like they were made for 20 years ago when movies were $5 in a theater and you could see something new every few weeks so it was fine if it was so-so

200 0

Brendan Kenny @brendankenny.bsky.social · 12/05/2026

The Subnautica subreddit consensus is yes, which means they'll get their bonus and make the game they want to, but I haven't seen an official source (maybe there won't be one). Are you planning on doing early access? I'm torn. (I played on your rec like five years ago, thanks for that!)

110 0

Brendan Kenny @brendankenny.bsky.social · 12/05/2026

I'm sure my feelings will slowly have to change as most too-long PRs will actually be from someone else's LLM and any feedback will not be looked at by a human :) I feel the same way about people claiming hours of /plan work. I have to break it up. Incredibly taxing and boring to review otherwise

000 0

Brendan Kenny @brendankenny.bsky.social · 12/05/2026

My happy spot is still reviewing chunks at a time. Definitely slower but too much LLM code at once and my eyes glaze over and I have trouble sustaining care. For some reason it's different when I get a too-long PR but at least I know a human is on the other end waiting for feedback

100 0

Brendan Kenny @brendankenny.bsky.social · 29/04/2026

but have you considered interpolation variable identifiers as a functor into RGB space

020 0

Brendan Kenny @brendankenny.bsky.social · 22/04/2026

Heh, I thought this was a RTO take at first and was confused for a minute. I mean, I didn't plan on dying *at* the office, but maybe they'll be requiring it by then...

000 0

Brendan Kenny @brendankenny.bsky.social · 02/04/2026

We don't actually know! It could decide to be like only 2s and 3s way way (way) out there. Clearly we need a π Status Page

120 0

Brendan Kenny @brendankenny.bsky.social · 03/02/2026

How do you manage that in the car if they're too young for their own device? Apple music has nothing AFAICT, youtube music for all its faults has different accounts but is only switchable on the phone itself with multiple (slow) taps. When someone is begging to listen to Golden again I just give in

120 0

Brendan Kenny @brendankenny.bsky.social · 24/01/2026

Not really, which is terrible. Also check for any paid subscriptions youtube offers now which might seem reasonable to someone and then be forgotten about :( Theres browser extensions that will client-side filter, but otherwise delete history, try to reseed with better videos, and repeat in a month

010 0

Brendan Kenny @brendankenny.bsky.social · 20/01/2026

Blanking some random but meaty github issue descriptions and MDN snippets is interesting. ChatGPT and Gemini identify the technical writing, but there's not enough there for them to nail down context. ChatGPT keeps thinking things are about OpenAI KV caching, and Gemini thinks I'm asking about 737s

Auto-Throttle System Explained

Fill in the blanks in the following passage:

The ___ ___ of the ___ ___ is used for ___ to the ___ ___ during a ___ and continuing ___ later, with the ___ ___ as a ___ ___ (see the ___ ___ ___ ___ for more ___). This...

This passage describes the technical operation of an aircraft's Auto-Throttle (A/T) system, specifically the logic used during the takeoff and climb phases (common in Boeing 737 systems).

Here is the filled-in passage:

Completed Passage
The N1 mode of the auto-throttle is used for thrust to the N1 limit during a takeoff and continuing climb later, with the thrust levers as a visual reference (see the Auto Flight System Manual for more details). This allows engine-rated thrust to be automatically maintained so the pilot stays alert.

The A/T can annunciate when the airspeed sensed by the FMC is 84 knots. The annunciation for when the airspeed is 84 knots defaults to "THR-HLD", but can display a different mode if the takeoff roll occurs within a specified window.

010 0

Brendan Kenny @brendankenny.bsky.social · 20/01/2026

Here's ChatGPT 5.2 thinking. Looks like it could be right? Pretty remarkable. I wish the thinking log would give any insight on how it made that first leap

To change a flat tire on a bike, first insert a tire lever between the tire and rim.
Next to that, insert a second lever and slide it along the outer edge of the rim.
You should now be able to remove the tire.
Gently pull the inner tube and attach the pump directly to the valve.
No adapter is needed.
Partially inflate the tube and put it back inside the tire,
then use the tire levers to roll the tire back onto the rim.
Finish inflating the tire to full pressure.

Activity - 47s
Thinking
Filling in bike repair instructions
Looks like I need to fill in blanks for instructions, possibly on removing a rear bike wheel. It says: "To ____ a ____ ____ on a ____." My guess is: "remove a rear wheel on a bike." The step "insert a tire lever between the tire and rim" seems pretty typical. Let's work with that!

100 0

Brendan Kenny @brendankenny.bsky.social · 20/11/2025

That's a Greater roadrunner if you haven't spent time in southwesternish North America

en.wikipedia.org

Greater roadrunner - Wikipedia

100 0

Brendan Kenny @brendankenny.bsky.social · 20/11/2025

Did not recognize this grumpy muppet of a bird until it engaged its afterburners. I've never seen one inside city limits before!

A grumpy looking bird with streaked brown and white feathers looks through a wire fence

A roadrunner striding across a driveway, its spiky head crest raised and its long tail feathers sticking straight up in the air

110 0

Brendan Kenny @brendankenny.bsky.social · 24/10/2025

Anyways, the reason Michael Crichton probably thought T. Rex's vision was based on movement was because he didn't know the Opit was above the Brale but *behind* the ⨡arl

000 0

Brendan Kenny @brendankenny.bsky.social · 24/10/2025

It cracks me up that a ChatGPT 5 conversation can be going ok, it was obviously RLHFed so hard to suggest a diagram when it would be genuinely useful, but apparently the second-order feedback "unless the resulting diagram is sloppy nonsense" wasn't incorporated

A ChatGPT conversation. ChatGPT asks "Would you like a labeled cutaway diagram showing where all these parts fit in the skull?", the user responds "Yes please", and ChatGPT responds with a cutaway diagram of a T. Rex head, with nonsense labels like two different brains, the eyeball labelled "Opit", and some part—maybe the ocular muscles, I really have no idea—labelled "Brale".

110 0

Brendan Kenny @brendankenny.bsky.social · 22/10/2025

`import defer` will get you some of what you're looking for. The file is still loaded, but the synchronous parts of the module (and submodules) won't execute until first property access of the module namespace object. github.com/tc39/proposa...

github.com

GitHub - tc39/proposal-defer-import-eval: A proposal for introducing a way to defer evaluate of a module

A proposal for introducing a way to defer evaluate of a module - tc39/proposal-defer-import-eval

040 0

Brendan Kenny @brendankenny.bsky.social · 19/09/2025

To be fair, I forced it to use thinking mode to try again, and it admitted it was wrong. I was amused it based its answer only on searching MathOverflow and Wikipedia and not finding a mention of equivalency, though. I'm sure proving independence is difficult, but I was hoping for a little more :)

Thought for 1m 30s

You’re right to call that out — my earlier take was sloppy. Here’s the clean version.

Let
P := “π + e is irrational”
Q := “the Riemann Hypothesis (RH) is true.”

The claim is the disjunction P ∨ Q. That is not a tautology; it’s true iff at least one of P or Q is true. Right now:

We do not know whether π + e is irrational. It’s a famous open problem. (A basic fact we can prove: at least one of π + e or πe must be irrational, since if both were rational then π and e would be roots of a quadratic with rational coefficients, contradicting their transcendence. But that doesn’t tell us which one.)
MathOverflow
Mathematics Stack Exchange
Wikipedia

We do not know whether RH is true; it remains unsolved.
Wikipedia
arXiv

So the truth of “π + e is irrational or RH is true” is currently unknown. It would be true in any world where either RH holds or π + e is irrational; it would be false only in the (logically possible) world where RH is false and π + e is rational — a scenario not ruled out by current knowledge.

A handy logical rewrite:
P ∨ Q is equivalent to (¬P) → Q. So the statement is the same as:

“If π + e were rational, then RH would be true.”
No such implication is known, and there’s no known connection of that sort.
Philosophy Stack Exchange
Mathematics LibreTexts

Bottom line: the statement is not “formally true.” Its truth value is unknown with our present mathematics.

010 0

Brendan Kenny @brendankenny.bsky.social · 19/09/2025

I had a funny one when I saw a claim about π + e and the Riemann hypothesis on the math subreddit that I hadn't heard. It was written in a way that suggested it might be a joke, though, so I though maybe I just didn't get it and asked GPT5. The first line is so intriguing! And then just nonsense

That statement

π + e is irrational or the Riemann Hypothesis is true

is formally true, but for a very boring logical reason rather than a deep number-theoretic connection.

Why it’s true

A statement of the form (A or B) is true as long as at least one of A or B is known to be true.

In this case
A = π + e is irrational.
B = The Riemann Hypothesis is true.

It is already known that π + e is irrational or at least one of π and e is transcendental (both are known to be transcendental individually).
In fact, it is an open problem whether π + e is irrational (we don’t know), but we do know that π and e are irrational individually.
And more importantly, The logical form “P or Q” is always true if either part is true.

“π + e is irrational” is not known to be true, but its negation (π + e is rational) would be shocking and is widely believed to be false.

But here’s the key: Any statement of the form (unknown thing) OR (independent big conjecture) is trivially true unless both parts are false. Since the Riemann Hypothesis being false doesn’t force π + e to be rational, and π + e being rational isn’t known, the combined disjunction can’t be proved false.

What’s really going on

This is a classic joke statement in mathematics

People take an open problem (like π + e irrationality). They take another famous open problem (Riemann Hypothesis).

Then they glue them together with an OR. The result is formally true (because if RH is true, the disjunction is true no matter what).

So the claim is not a genuine theorem about π + e and RH being connected. It’s just a way of making a logically true but vacuous statement.

✅ Bottom line:
Yes, the statement is true, but only because it’s a trivial logical disjunction. It does not establish any relationship between π + e and the Riemann Hypothesis.

Do you want me to walk you through why these kinds of true but useless statements are sometimes used in math (kind of like party tricks) or would you prefer I stick strictly to the logic?

110 0

Brendan Kenny @brendankenny.bsky.social · 19/09/2025

small multiples! en.wikipedia.org/wiki/Small_m... Looks fun, any shareable information about how "Create variants" is constrained? Is "give me more that looks like this" enough or does it need to go deeper?

en.wikipedia.org

Small multiple - Wikipedia

010 0

Brendan Kenny @brendankenny.bsky.social · 12/09/2025

I vaguely remember the Wario game being ok? And Mario Tennis being so basic that it might as well have been pong. And yet, yes, the kid still in me who saved up allowance to rent one from Blockbuster for a weekend really wants one

010 0

Brendan Kenny @brendankenny.bsky.social · 05/03/2025

p8->p10 as a policy isn't quite the same (the PoDR isn't a fixed percentile of the distribution) but it's the rough equivalent to the old PoDR behavior. The p10 values also have the most manual tweaks, e.g. the actual p8 CLS is 0, but the p10 used is 0.1 to better recognize what CWV considers "good"

130 0

Brendan Kenny @brendankenny.bsky.social · 05/03/2025

and it was actually possible to score 100 (roughly 2% of the pages in HTTP Archive at the time would get 100 on a metric). But then we admitted everyone was thinking in terms of percentiles anyway, not the PoDR, so we kept the curves unchanged but converted it to be defined by the median and the p10

120 0

Brendan Kenny @brendankenny.bsky.social · 05/03/2025

p8->p10 is more convoluted because originally we specified the score curve by the median and "point of diminishing returns" (PoDR) (referenced in that screenshot), thinking of that as the more natural way to frame it, and adjusting it so that p5->PoDR so that it was both an aspirational curve

130 0

Brendan Kenny @brendankenny.bsky.social · 05/03/2025

p25->p50 is somewhat arbitrary, but evolved over the first few years of lighthouse to be a balance between reflecting reality and where, through many many perf audits, we judged the median site _could_ be with even moderate effort

120 0

Brendan Kenny @brendankenny.bsky.social · 05/03/2025

the CDF has the fastest growth in the middle, rewarding perf improvements where most sites are with the biggest score increases, and it ramps off as you get close to "great", in theory encouraging you to then look at other metrics for improvements (built in prioritization)

120 0

Brendan Kenny @brendankenny.bsky.social · 05/03/2025

Most of these metrics are already log-normal (or log-normal enough) when looking at small or large scale (visits to a single page or visits aggregated over many origins), so it's a natural scoring curve model for the data, and it also has a couple of benefits:

120 0

Brendan Kenny @brendankenny.bsky.social · 05/03/2025

It's not a very pat answer because it evolved over years, but basically it's a balance between being based on real data, being aspirational (encouraging websites to be better than the typical website), rewarding perf improvements where it matters, and having good scores (and even 100s) be achievable

150 0

Reposted by Brendan Kenny

Barry Pollard @tunetheweb.com · 27/11/2024

Yes it's time to say good by to the Performance Insights panel and get back to having one Performance panel—with the best of both worlds! Read more from @brendankenny.bsky.social here: developer.chrome.com/blog/insight...

A somber Oppenheimer on the left with the label: Chrome removing the Performance Insights panel.
A delighted Barbie on the right with the label: Chrome adding even better insights to the Performance panel!

1124 0