Rob Whelan 🌱
@robw.bsky.social
Turns out I can skip posts not intended for me. AI/ML @ Grafana; everything is UX; let's make it better; he/him
Landlord (desperately): but surely, if the good I do in the world with the rent money exceeds the harm ICE inflicts, then—
Grimdark superhero [holding landlord over roof edge]: and... does it? DOES IT?!
Grimdark superhero [holding landlord over roof edge]: and... does it? DOES IT?!
November 11, 2025 at 3:21 PM
Landlord (desperately): but surely, if the good I do in the world with the rent money exceeds the harm ICE inflicts, then—
Grimdark superhero [holding landlord over roof edge]: and... does it? DOES IT?!
Grimdark superhero [holding landlord over roof edge]: and... does it? DOES IT?!
Thx for this thread!
We must, we must. I do feel like a big part of my job is getting devs thinking more (& more) about what it's *really* like to work with a product/feature you didn't build, don't trust or understand; to learn internal platform, so many things + diff experiences we can change
We must, we must. I do feel like a big part of my job is getting devs thinking more (& more) about what it's *really* like to work with a product/feature you didn't build, don't trust or understand; to learn internal platform, so many things + diff experiences we can change
November 9, 2025 at 6:55 PM
Thx for this thread!
We must, we must. I do feel like a big part of my job is getting devs thinking more (& more) about what it's *really* like to work with a product/feature you didn't build, don't trust or understand; to learn internal platform, so many things + diff experiences we can change
We must, we must. I do feel like a big part of my job is getting devs thinking more (& more) about what it's *really* like to work with a product/feature you didn't build, don't trust or understand; to learn internal platform, so many things + diff experiences we can change
We ride after elevenses, but then stop again immediately to make luncheon arrangements
November 9, 2025 at 12:33 PM
We ride after elevenses, but then stop again immediately to make luncheon arrangements
Vague idea is that dramatic exaggeration like that is being silly, and I'll accept it in silly mode (which is fine)
if they actually want to be serious, then we should talk in real terms about what's really happening
(Dunno if my kids actually get this lesson, but it's what I'm trying for :D)
if they actually want to be serious, then we should talk in real terms about what's really happening
(Dunno if my kids actually get this lesson, but it's what I'm trying for :D)
November 9, 2025 at 10:58 AM
Vague idea is that dramatic exaggeration like that is being silly, and I'll accept it in silly mode (which is fine)
if they actually want to be serious, then we should talk in real terms about what's really happening
(Dunno if my kids actually get this lesson, but it's what I'm trying for :D)
if they actually want to be serious, then we should talk in real terms about what's really happening
(Dunno if my kids actually get this lesson, but it's what I'm trying for :D)
I think this is where I'd fake-panic; omg, you must be near starvation, but what if there's not even any food in the house at all?? We'd better search high & low — aren't crayons edible?? Can we hunt squirrels in the yard?? And then a mad hunt turns up dinner at the table, definitely magic
November 9, 2025 at 10:54 AM
I think this is where I'd fake-panic; omg, you must be near starvation, but what if there's not even any food in the house at all?? We'd better search high & low — aren't crayons edible?? Can we hunt squirrels in the yard?? And then a mad hunt turns up dinner at the table, definitely magic
I do worry about the path to senior dev — I leaned to manage complexity etc over years of other engineers reviewing my work, saying "yes, that's a fix BUT it'll confuse the dev after you", "this'll break silently" etc
but AI seems much more valuable for sr devs than jrs; what's the growth path now?
but AI seems much more valuable for sr devs than jrs; what's the growth path now?
November 9, 2025 at 9:15 AM
I do worry about the path to senior dev — I leaned to manage complexity etc over years of other engineers reviewing my work, saying "yes, that's a fix BUT it'll confuse the dev after you", "this'll break silently" etc
but AI seems much more valuable for sr devs than jrs; what's the growth path now?
but AI seems much more valuable for sr devs than jrs; what's the growth path now?
Very true; LLMs are happy to accelerate me into horrific tech debt — e.g. I ask for [big thing A] then try to prompt it into [big thing B] without stopping to understand & refactor.
It's still my job to manage complexity & maintainability — if I do that, it's great.
(But *only* if I do that well.)
It's still my job to manage complexity & maintainability — if I do that, it's great.
(But *only* if I do that well.)
November 9, 2025 at 9:07 AM
Very true; LLMs are happy to accelerate me into horrific tech debt — e.g. I ask for [big thing A] then try to prompt it into [big thing B] without stopping to understand & refactor.
It's still my job to manage complexity & maintainability — if I do that, it's great.
(But *only* if I do that well.)
It's still my job to manage complexity & maintainability — if I do that, it's great.
(But *only* if I do that well.)
Amazing. He could use a different keyboard every day for a year, as long as he has 12 weeks PTO. Or he could loop back to the start, and it's still a change every day, but that'd work with just 2 keyboards, and that's not special. Point is, he may need an additional 165 keyboards just to be sure
November 8, 2025 at 10:21 PM
Amazing. He could use a different keyboard every day for a year, as long as he has 12 weeks PTO. Or he could loop back to the start, and it's still a change every day, but that'd work with just 2 keyboards, and that's not special. Point is, he may need an additional 165 keyboards just to be sure
People with harmful bias... train LLMs with harmful bias
youtu.be/Y-Elr5K2Vuo
youtu.be/Y-Elr5K2Vuo
"I Learned It By Watching You" Anti-Drug PSA
YouTube video by ticklemeozmo
youtu.be
November 7, 2025 at 12:09 PM
People with harmful bias... train LLMs with harmful bias
youtu.be/Y-Elr5K2Vuo
youtu.be/Y-Elr5K2Vuo
Organizationally, it just makes sense for them to have a dedicated division
And a "VP, Scams & Banned Goods" leadership role at least, but potentially a Chief Scam Officer
or Head of Scam?
And a "VP, Scams & Banned Goods" leadership role at least, but potentially a Chief Scam Officer
or Head of Scam?
November 7, 2025 at 11:58 AM
Organizationally, it just makes sense for them to have a dedicated division
And a "VP, Scams & Banned Goods" leadership role at least, but potentially a Chief Scam Officer
or Head of Scam?
And a "VP, Scams & Banned Goods" leadership role at least, but potentially a Chief Scam Officer
or Head of Scam?
& abstracly "we've figured this out & can stop debating it" cannot be true unless "we" is the same as yesterday
(and *that* won't happen as long as new humans keep being born & joining the chat)
(and *that* won't happen as long as new humans keep being born & joining the chat)
November 7, 2025 at 11:51 AM
& abstracly "we've figured this out & can stop debating it" cannot be true unless "we" is the same as yesterday
(and *that* won't happen as long as new humans keep being born & joining the chat)
(and *that* won't happen as long as new humans keep being born & joining the chat)
A lot going on as usual, but also really weird things happen when people learn to enforce "the rules" rather than learn how to make life good
Straw Hat Riot - Wikipedia
en.wikipedia.org
November 5, 2025 at 7:04 AM
A lot going on as usual, but also really weird things happen when people learn to enforce "the rules" rather than learn how to make life good
My partner's first novel was in Anthropic's training data (and I support the class action suit) & I feel strongly that AI needs careful regulation.
But basically, the reality is messy (as it often is), and we should stay connected to it. That's my only point here.
But basically, the reality is messy (as it often is), and we should stay connected to it. That's my only point here.
November 2, 2025 at 10:09 PM
My partner's first novel was in Anthropic's training data (and I support the class action suit) & I feel strongly that AI needs careful regulation.
But basically, the reality is messy (as it often is), and we should stay connected to it. That's my only point here.
But basically, the reality is messy (as it often is), and we should stay connected to it. That's my only point here.
Right, it's a mistake to treat it as deterministic or reliable. I do not do that.
I'm deciding if the tech is valuable based on live services that use it (with other tech).
A lie is an attempt to deceive — LLMs are definitely not doing this. They know nothing about the correctness of their output.
I'm deciding if the tech is valuable based on live services that use it (with other tech).
A lie is an attempt to deceive — LLMs are definitely not doing this. They know nothing about the correctness of their output.
November 2, 2025 at 10:03 PM
Right, it's a mistake to treat it as deterministic or reliable. I do not do that.
I'm deciding if the tech is valuable based on live services that use it (with other tech).
A lie is an attempt to deceive — LLMs are definitely not doing this. They know nothing about the correctness of their output.
I'm deciding if the tech is valuable based on live services that use it (with other tech).
A lie is an attempt to deceive — LLMs are definitely not doing this. They know nothing about the correctness of their output.
I understand lots of pushback to LLMs, especially using them for creative work or fact lookup;
the tech is very powerful as a translation & coordination layer between people & non-LLM tools, though.
Framing it as wholly useless (& anthropomorphizing it — people "lie", algorithms don't) feels off.
the tech is very powerful as a translation & coordination layer between people & non-LLM tools, though.
Framing it as wholly useless (& anthropomorphizing it — people "lie", algorithms don't) feels off.
November 2, 2025 at 9:47 PM
I understand lots of pushback to LLMs, especially using them for creative work or fact lookup;
the tech is very powerful as a translation & coordination layer between people & non-LLM tools, though.
Framing it as wholly useless (& anthropomorphizing it — people "lie", algorithms don't) feels off.
the tech is very powerful as a translation & coordination layer between people & non-LLM tools, though.
Framing it as wholly useless (& anthropomorphizing it — people "lie", algorithms don't) feels off.
It's safe! Auto-fixes for minor errors (after confirming with you). Good test coverage and worked great in our dev envs with sample docs
Oh, wait, a full book manuscript... yeah, it's going to OOM and hang, and then auto-save your file with zero bytes
Oh, wait, a full book manuscript... yeah, it's going to OOM and hang, and then auto-save your file with zero bytes
November 2, 2025 at 7:27 PM
It's safe! Auto-fixes for minor errors (after confirming with you). Good test coverage and worked great in our dev envs with sample docs
Oh, wait, a full book manuscript... yeah, it's going to OOM and hang, and then auto-save your file with zero bytes
Oh, wait, a full book manuscript... yeah, it's going to OOM and hang, and then auto-save your file with zero bytes
100%, this is a better situation!
November 1, 2025 at 10:35 PM
100%, this is a better situation!
You took the time to write a long, irritated, insulting reply to a social media post.
It's worth considering that maybe the post was just not meant for you, and selecting "Show less like this" for it.
It's worth considering that maybe the post was just not meant for you, and selecting "Show less like this" for it.
October 28, 2025 at 12:57 PM
You took the time to write a long, irritated, insulting reply to a social media post.
It's worth considering that maybe the post was just not meant for you, and selecting "Show less like this" for it.
It's worth considering that maybe the post was just not meant for you, and selecting "Show less like this" for it.
Excellent, fully prepared for casting cat treats before cats (who will appreciate them) or pearls before swine (who will not)
whatever dramatic gesture your conversations call for, you're ready
whatever dramatic gesture your conversations call for, you're ready
October 26, 2025 at 9:18 PM
Excellent, fully prepared for casting cat treats before cats (who will appreciate them) or pearls before swine (who will not)
whatever dramatic gesture your conversations call for, you're ready
whatever dramatic gesture your conversations call for, you're ready
I still know the phone number of my little sister's best friend — 764-4462 — because at maybe 5 years old she was so excited to be using the phone like a big kid and talked about it constantly 😄
October 13, 2025 at 9:47 PM
I still know the phone number of my little sister's best friend — 764-4462 — because at maybe 5 years old she was so excited to be using the phone like a big kid and talked about it constantly 😄
Classic "omnipotence" paradox — can Dog dig a hole so deep even He cannot get out if it
October 12, 2025 at 8:10 AM
Classic "omnipotence" paradox — can Dog dig a hole so deep even He cannot get out if it
Not-so-subtle dig at Americans, I believe!
October 9, 2025 at 11:56 AM
Not-so-subtle dig at Americans, I believe!
"Maybe it's Maybelline, ...maybe it's Moraxella lacunata!" as the jingle goes
Moraxella lacunata - Wikipedia
en.wikipedia.org
October 9, 2025 at 11:22 AM
"Maybe it's Maybelline, ...maybe it's Moraxella lacunata!" as the jingle goes
No, it's extremely bad, and I think it's worth figuring out what works to stop this.
I don't think you should throw your life away if the result were that they can justify sending in military and escalating violence
but if the result were really reducing raids, I'd be listening
I don't think you should throw your life away if the result were that they can justify sending in military and escalating violence
but if the result were really reducing raids, I'd be listening
October 4, 2025 at 6:31 PM
No, it's extremely bad, and I think it's worth figuring out what works to stop this.
I don't think you should throw your life away if the result were that they can justify sending in military and escalating violence
but if the result were really reducing raids, I'd be listening
I don't think you should throw your life away if the result were that they can justify sending in military and escalating violence
but if the result were really reducing raids, I'd be listening
There are a couple comments like this.
How does "defend yourself" work in a small apartment with your vulnerable family, when your door is broken down by dozens & dozens of armed men? & it mentions they used flashbang grenades.
If you manage to shoot some of them, what happens next?
How does "defend yourself" work in a small apartment with your vulnerable family, when your door is broken down by dozens & dozens of armed men? & it mentions they used flashbang grenades.
If you manage to shoot some of them, what happens next?
October 4, 2025 at 6:20 PM
There are a couple comments like this.
How does "defend yourself" work in a small apartment with your vulnerable family, when your door is broken down by dozens & dozens of armed men? & it mentions they used flashbang grenades.
If you manage to shoot some of them, what happens next?
How does "defend yourself" work in a small apartment with your vulnerable family, when your door is broken down by dozens & dozens of armed men? & it mentions they used flashbang grenades.
If you manage to shoot some of them, what happens next?