maggieappleton.com
Design engineer exploring applied AI and hacky prototypes.
Adores digital gardening, end-user development, and embodied cognition. Makes visual essays about design, programming, and anthropology.
đ London
đą maggieappleton.com
443 posts
17,415 followers
479 following
Regular Contributor
Active Commenter
comment in response to
post
Just witnessed poor quality debate
bsky.app/profile/magg...
comment in response to
post
I appreciate your existence though. It motivates me to keep working on responsible and transparent LLM interface design, grounding models in high quality evidence, and making model reasoning clear and understandable to users.
Thanks. Otherwise we'll all drown in the weak sauce critique.
comment in response to
post
This type of blanket critique of LLMs by offering only the most obviously wrong outputs by the worst models, while assuming no human ever does due diligence or evaluates the truthfulness of claims they read on the web, is weak sauce.
You're why I have to post the extremely boring caveat EVERY TIME
comment in response to
post
Google's AI results are notoriously shit in a way ChatGPT and Claude are not, and as I said in my follow up, I obviously use basic common sense and double check facts / stats.
LLMs are invaluable as long as you're not an idiot and deploy your critical thinking skills.
bsky.app/profile/magg...
comment in response to
post
Oh I had not! Wouldn't have thought of that. Will give it a try.
Sadly a lot of local community events still don't have much of an online presence though. At least small scale, non-commercial ones: parenting groups, workout classes, etc. Lots just run on community noticeboards. Quaint in 2025.
comment in response to
post
There is a shocking amount to learn once you bring a baby home, and most of it is not going to come to you through intuition.
There's only so much community / NHS support available, and ChatGPT fills in a thousand tiny gaps in my knowledge and lets me ramble at it for hours to troubleshoot issues.
comment in response to
post
Boring caveats: Yes, of course I double check important statistics and facts. Web search is built into most model interfaces and I have nothing else to do at 3am but carefully check the source links.
I can't wait for the day where I don't have to post this boring caveat on anything re: LLMs.
comment in response to
post
Iâve also found it invaluable for this use case. I fall into a weird tax category (US/UK dual citizen) that makes everything complicated. Accuracy is good when you upload official government rules and guideline PDFs as context and have it walk you through tax policies and calculations step-by-step.
comment in response to
post
I went to this as a database muggle and learned tons from the talks. Good case of being thrown in the deep end. Even if I couldn't grok the more technical implementation details, all the speakers framed problems well and made me care on a philosophical level.
TLDR No previous knowledge required.
comment in response to
post
Lol, itâs so funny and clarifying from the other side what is actually useful!
Ditto on the heating pad. I use a hot water bottle and itâs one of my most valued possessions. Cures almost everything.
comment in response to
post
Yeah no earplugs are going to block out the volume level of a hungry baby, ha.
I have Loops which are designed to reduce noise vs. block out out and they work perfectly. Can still hear major squeaks, grunts, and cries - even when sleeping - but at a less acute and panic-inducing level.
comment in response to
post
Yes, also a good one! A few people gifted us subscriptions to healthy ready meal companies. I did a ton of bulk cooking ahead of time to load up the freezer which I don't regret at all.
comment in response to
post
Oh yeah that's a nice indulgent option. Good for a gift!
My tea definitely tastes worse/different out of a thermos. But I also like having it covered. Slightly terrified of spilling hot liquid over the baby đŹ
comment in response to
post
Theyâre good at explaining how small adjustments will affect the outcome. E.g. I can ask how to make brownies fudgier, more cake-like, vegan, GF. Theyâve read enough baking theory online to synthesise the principles behind it well.
And can suggest ingredient swaps if youâre out of something.
comment in response to
post
I think of this as âhigher floors and higher ceilingsâ.
Weâll have higher standards for standard, everyday content (and everything that falls below it gets dismissed as slop/not worth our time).
And then higher ceilings for what we expect the best people to produce.
comment in response to
post
Oh yes I use them for that a ton. But for the final words - the bit where you have to find the exact, concise phrases that say the thing you want to say, in the right narrative tone, with the right structure to them - theyâre sadly not much help.
comment in response to
post
I do some version of this too. Very heavy use of Whisper for voice input. But itâs always just initial conversations, research directions, and Socratic questioning.
Once it comes to writing the final words, I find LLMs canât help that much. Other than as an expensive thesaurus.
comment in response to
post
I canât use ChatGPT to write because it doesnât know what the fuzzy, half-formed ideas in my head are that Iâm trying to crystallise into words.
Until we hook LLMs up to telepathy machines itâs not much use.
But itâs a good rubber duck to talk to about those ideas as a preliminary step.
comment in response to
post
I hope this includes eating my baby's body weight in chocolate.
comment in response to
post
Strong agree!
The most forgettable talks are always âhow to do X in Y framework.â No one will care in 6 months.
Sad I missed All Day Hey. Have heard good things about it and will try to come next year.
comment in response to
post
Now ÂŁ30 for a subscription. Covers 2 years, but if you have multiple kids youâd have to renew at some point.
My vibe coded clone wonât be as good, but itâll do the basics of warning me when a growth stage is coming up.
comment in response to
post
Agree it would be more helpful to focus on macro, corporation-scale energy usage and policies.
But I see a lot of moral high ground, personal-level "you shouldn't use ChatGPT for your daily life" takes. As if refusing to get an LLM to adjust a brownie recipe for you is an act of climate justice.
comment in response to
post
Trying to make a hacky version of a baby growth tracker. Apps that tell you when your baby is in a growth spurt or about to acquire new skills.
There are a couple popular ones like The Wonder Weeks out there, but they are cashing in on the parent ÂŁÂŁÂŁ and Iâm done paying for baby stuff, lol
comment in response to
post
Yes this is the real benefit. Total control! Any app where the design is slightly shit or includes ads gets replaced by a vibe coded version immediately.
comment in response to
post
A close approximation of my process as well.
comment in response to
post
Seems like a fair price to pay for everyone maintaining their manners muscles
comment in response to
post
Thank you! Already getting more sleep and enjoying it more every day. The snuffle noises only get better.
comment in response to
post
Knowledge management generally. I keep references and notes in there for writing and bigger projects.
Some tasks go in there but only if not urgent and related to a specific project. Day-to-day tasks (groceries, chores, obligations) go in Todoist which is better designed for tasks on the go.
comment in response to
post
Oh yay! đ Congratulations! Let me know if you need anything or want solicited advice.
I found it super helpful to have women around me who had just been through the process and could answer Qâs on all the small, boring things; induction, swaddles, milk supply, etc.
comment in response to
post
Ended up hard exiting out of the probability graph. Went to 41 weeks and 4 days before the midwives recommended being induced for babyâs safety. But then baby was in a bit of distress so we did an unplanned c-section.
Thankfully no complications and all went fine đ
comment in response to
post
But also true that you become like the Grinch whose heart expands three sizes.
Seems completely fine to never sleep and attend to this tiny creatureâs every need because theyâre the best thing on earth and your life was a soulless, lifeless void before their existence.
comment in response to
post
Things I found stressful in my previous life seem like a joke. Design review went badly. Train delayed an hour. Landlord being difficult. Taxes need filing.
Now doing these while carrying a baby who needs me to survive, must be fed/changed every hour, and screams bloody murder if not attended to.
comment in response to
post
Yes, all here: maggieappleton.com/tools-though...
comment in response to
post
Congratulations!
comment in response to
post
Oh yeah the dataset over covers 37 weeks pregnant onwards so canât predict dates earlier than that
comment in response to
post
I never understand the point of these efforts. Whatâs the end goal? Eventually DM folks with crypto recommendations?
Step 1. Impersonate person
Step 2. âŚ..?
Step 3. PROFIT!
comment in response to
post
Also that stair machine comes in handy for side stepping when youâre trying to get labour moving đââď¸ wish gyms had more flexible options for preggos, but I suppose weâre a small and weird audience to serve lol
comment in response to
post
Oh I found it through the mailing list. The Substack one. Maybe she has multiple?