| Dec 24 |
@fractaledmind One of my 2026 goals was to “lean into html on Twitter”. I fear you may have beaten me to it 😅
|
2 |
0 |
0 |
0 |
109 |
. |
| Dec 22 |
From our team's Granola Wrapped 😅
I may have finally arrived at the lifestyle I set out to build. https://t.co/rg6kWb6zVa
|
4 |
0 |
0 |
212 |
123 |
. |
| Dec 21 |
"Don't underthink it" is an underused mantra.
Sit down, write down your thoughs, create the spreadsheet, run the numbers. Untangle what makes sense from what doesn't. Go deeper than surface level. Have good answers for yourself and others.
|
2 |
0 |
1 |
245 |
240 |
. |
| Dec 20 |
I keep seeing these charts of engineering task that take LLMs 4.5 hours to complete, but no one ever says what the task is or if it’s things that are actually useful.
|
3 |
0 |
0 |
434 |
166 |
. |
| Dec 20 |
The future of frontend engineering is in having in-depth knowledge of CSS, not having in-depth knowledge of javascript.
|
0 |
0 |
0 |
146 |
119 |
. |
| Dec 17 |
@fractaledmind @rikschennink Tbh I don't disagree, I just haven't had a need for it since I stopped using it.
|
0 |
0 |
0 |
0 |
109 |
. |
| Dec 17 |
@fractaledmind @rikschennink Gotcha! Maybe one day when I ditch my puritanistic ways and go back to having a build step 😅 enjoyed your most recent thread
|
1 |
0 |
0 |
0 |
153 |
. |
| Dec 17 |
@fractaledmind @rikschennink I personally love the simplicity of just throwing an svg in my icons folder and using inline_svg to reference it
|
0 |
0 |
0 |
0 |
141 |
. |
| Dec 17 |
@fractaledmind @rikschennink Love this generally. Am I correct in saying with this solution when you add a new icon you have to make a new class for it?
|
0 |
0 |
0 |
0 |
152 |
. |
| Dec 16 |
@jackmcdade Did this last weekend and the best collection is
https://t.co/AKkOJxwaE7
|
5 |
1 |
0 |
0 |
85 |
. |
| Dec 13 |
@Jivansh77 @karpathy @karpathy if you’re interested in this check out https://t.co/EgIFQGIGpH - a treatise against the 500lb website
|
61 |
7 |
0 |
0 |
132 |
. |
| Dec 12 |
But I thought you *needed* to write everything in typescript to not get left behind by AI assisted coding?
|
2 |
0 |
0 |
404 |
106 |
. |
| Dec 06 |
@adamwathan Any chance you have a blog post link that covers what you’d include in “all of the accessibility stuff”? Also at what % coverage would you consider the new browser features to be acceptable here? Popovertarget is at 88% coverage but doesn’t get you keyboard navigation
|
0 |
0 |
0 |
0 |
280 |
. |
| Dec 06 |
This is true if…
- You are building very highly interactive frontends
- You don’t want to spend any time figuring out a playbook/set of patterns (which is completely fair and true of 99% of people)
It’s not because of any fundamental difficulty with doing this stuff without js, It’s just that js libs have consumed everything so there’s not as much material out there.
Because of this I don’t agree with the “you’ve got to be a masochist” framing. You just want to accept more up front work for lower long term complexity
|
5 |
0 |
1 |
923 |
525 |
. |
| Dec 01 |
IMO "Every meeting must have an agenda" is a sign of bad culture. If your people aren't smart enough not to schedule unnecessary meetings, or not to have discipline with who they invite & respect for others' time, you have bigger problems. https://t.co/XrwJQeDAni
|
2 |
0 |
0 |
198 |
267 |
. |
| Dec 01 |
https://t.co/qP1EWW4AqJ
|
1 |
0 |
0 |
185 |
23 |
. |
| Nov 28 |
@ze_rusty Plz stop grifting & causing concern https://t.co/8h6kQM88qM
|
11 |
0 |
0 |
0 |
73 |
. |
| Nov 28 |
This is the best browser agent right now btw. Working with the guys on some very fun stuff, useful browser agents *might* just become a thing soon.
|
1 |
0 |
0 |
339 |
147 |
. |
| Nov 27 |
@hunvreus From what I’ve observed people who like parallel agents work more in codebases that require a lot of boilerplate and where backend/frontend is separate, haven’t seen any rails folks talking about them for example
|
1 |
0 |
0 |
0 |
222 |
. |
| Nov 27 |
@StartupArchive_ @george__mack great one for the high agency list
|
0 |
0 |
0 |
0 |
65 |
. |
| Nov 26 |
Claude being very poetic this morning https://t.co/Zs47OzKZtZ
|
2 |
0 |
0 |
270 |
61 |
. |
| Nov 25 |
@amya_wilks Is the thesis right? I can see current positioning is beeper alternative but I’m guessing you’re going for “An actually-good Siri hooked up to your contacts” or similar? Just signed up, excited to try the beta!
|
0 |
0 |
0 |
0 |
222 |
. |
| Nov 25 |
Eyyy
https://t.co/Ir3d3Havqz
|
2 |
0 |
1 |
308 |
29 |
. |
| Nov 25 |
Met a guy at a wedding last week who had read this book. It had created a completely closed epistemic loop, where every point I made was returned with either “We don’t know that because AI is grown not built” or “That is just blind faith”. One of the most frustrating conversations with a stranger I’ve ever had. I say this as someone who is way-more-open-than-most to “doomer” perspectives.
I’m guessing that was one of the goals of the authors - to create an ideology that withstands critique, so hats off to them for that. 100% brain-wormed. Sample of one of course. Interested to hear others experiences
|
4 |
0 |
0 |
1.1k |
608 |
. |
| Nov 24 |
@zebriez The most succinct I got to was “Professional theory of mind”, bit of a mouthful though
https://t.co/ugz6QrDY7I
|
3 |
0 |
0 |
0 |
121 |
. |
| Nov 22 |
If you had told me when I tweeted this that this still wouldn't be mainstream in 2025 I would have been surprised. Lots of progress, but we're still pretty safe and not too far from where we were 2 years ago.
|
3 |
0 |
1 |
425 |
208 |
. |
| Nov 22 |
@jacksharkey11 Yes sir, this is the correct life motto
https://t.co/mfI4i7o7Vh
|
0 |
0 |
0 |
0 |
79 |
. |
| Nov 21 |
Amazing how the narrative brain works.
The way Ryanair works is the first seats are cheap and the prices go up as the cheap ones are sold. This is the only information you need to explain this phenomenon, and yet…
|
2 |
0 |
0 |
273 |
214 |
. |
| Nov 21 |
@thatguybg Genuinely a banger
|
1 |
0 |
0 |
0 |
29 |
. |
| Nov 21 |
🤣
|
2 |
0 |
0 |
362 |
1 |
. |
| Nov 20 |
@RaulOnRails @shl Solely GitHub yup, apparently they trained on a bunc of labelled data but I’d guess you could do a decent job with a decent prompt
|
0 |
0 |
0 |
0 |
148 |
. |
| Nov 20 |
@RaulOnRails @shl More like this https://t.co/oTKSaQevya
|
1 |
0 |
0 |
0 |
56 |
. |
| Nov 20 |
@RaulOnRails @shl Thanks! Not exactly what I'm looking for. Weave has a feature were they calculate an "output score" for every commit/PR, which his based on effort & impact - use case for me isn't daily check in - it's end of week/month seeing if we focused on high impact work https://t.co/QEwbfRrO4C
|
1 |
0 |
0 |
0 |
306 |
. |
| Nov 20 |
Linear is a tool that works well for a certain type of team, which is IMO <5% of software teams.
Most teams should not use Linear.
|
0 |
0 |
0 |
163 |
135 |
. |
| Nov 19 |
Is anyone building basic AI-powered engineering analytics (scan commits, figure out how much effort/impact is required, give it a score, see charts on per developer basis) that's not exorbitantly expensive?
|
3 |
0 |
1 |
347 |
206 |
. |
| Nov 18 |
The most important trait for hiring or getting hired is Professional Theory of Mind. If you can intuit what matters to other people (customers, managers, clients), you need only moderate management & can figure the rest out, but if you can't, you become a burden
|
6 |
0 |
0 |
2k |
266 |
. |
| Nov 16 |
There will be movies made about this werk and this man. You couldn’t script it, incredible scenes
|
3 |
0 |
0 |
793 |
97 |
. |
| Nov 11 |
https://t.co/a6WRRxVca6
|
2 |
0 |
0 |
120 |
23 |
. |
| Nov 11 |
Many reasons for this - one under discussed one is that there’s a pretty bad “kind truth vs unkind lie” problem in Ireland because everyone knows each other and no one wants to be known as the dick, which can only be fully escaped by leaving.
|
3 |
0 |
1 |
303 |
242 |
. |
| Nov 11 |
@prisencotech @htmx_org Geo replication for both app&db is a much heavier lift (maintenance & cost wise) than a shimmer effect and link activation though.
|
0 |
0 |
0 |
0 |
162 |
. |
| Nov 11 |
@prisencotech @htmx_org My solution to this isn’t a js framework btw, but I’d reach for hx-indicator adding a shimmer effect to the main div, and another js snippet that gives the active class to the clicked link, before id reach for view transitions
|
2 |
0 |
0 |
0 |
250 |
. |
| Nov 11 |
@prisencotech @htmx_org I agree 100% with that last statement, ultimately it comes down to how it feels. If my app server takes 20ms to return the page, but when I test on hotel WiFi in SF and server’s in Europe, if there’s any latency at all (there will be) it feels slower than SPAs, and that’s where…
|
2 |
0 |
0 |
0 |
303 |
. |
| Nov 11 |
@htmx_org I think view transitions are less exciting than they seem because they have to wait for the network so they always feel slow, they were one of the only disappointments in my research for the bsdc talk, kind of a red herring for now at least
|
2 |
0 |
0 |
0 |
250 |
. |
| Nov 11 |
@WisprFlow I can't get the fn key on my Logitech keyboard to trigger Wispr Flow, is that a known issue?
This is what Claude's telling me https://t.co/YQLp7vExFW
|
0 |
0 |
0 |
0 |
160 |
. |
| Nov 11 |
There's a massive opportunity to build an AI-powered messaging app on top of the Matrix protocol.
So much context can be inferred from chat, and sending & staying on top of dms is an actually-useful AI use case.
|
2 |
0 |
1 |
313 |
217 |
. |
| Nov 10 |
@ryandavogel This is sick. Can I use it for personal email too?
|
0 |
0 |
0 |
0 |
63 |
. |
| Nov 09 |
@Liv_Boeree @r00k @spicyliltoaster Really? Nurturer + health-conscious + optimist + funny + east coast + high-openness - they don't seem like traits that are anti-correlative. What makes you think that's a unicorn? genuinely curious https://t.co/3skv6QbMwV
|
0 |
0 |
0 |
0 |
256 |
. |
| Nov 09 |
🤦♂️ https://t.co/Fmg7LQL1nZ
|
1 |
0 |
0 |
155 |
28 |
. |
| Nov 09 |
Them: "Which do you prefer? Alpine, react, or stimulus?"
Me: "Yes"
|
1 |
0 |
0 |
337 |
67 |
. |
| Nov 09 |
This is the sweet spot for me when building components with more-than-basic interactivity: A Minijs component that uses preact standalone for templating.
- Easy to follow: Very little framework-specific knowledge needed.
- I can do simple slot-based composition in the dom. e.g reorder/remove parts.
- I can use jsx for templating (technically htm but looks and feels the same.
- Super performant
- No build step
|
3 |
1 |
0 |
720 |
416 |
. |
| Nov 08 |
Seeing this "AI written smell" everywhere now. Just me? https://t.co/xdGloqHoQK
|
2 |
0 |
1 |
246 |
79 |
. |
| Nov 08 |
I had no idea Ireland was this far detached from the rest of Europe when it comes to fertility.
|
1 |
0 |
0 |
288 |
95 |
. |
| Nov 07 |
@lottsnomad 😂 didn't realize the gender ratio. Looks like The Bachelorette
|
2 |
0 |
0 |
0 |
74 |
. |
| Nov 07 |
@michaelque22 Various companies that handle this. Adapty will give you the money up front for 1% interest if you use their platform. You switch the bank account on App Store Connect, they pay you up front, the apple payout goes to them and they forward anything remaining to you
|
4 |
0 |
0 |
0 |
278 |
. |
| Nov 06 |
@iamgdsa Love market-product fit
|
1 |
0 |
0 |
0 |
32 |
. |
| Nov 06 |
@ChrisEsplin @ryanflorence Calling it Edge Homes is diabolical
|
1 |
0 |
0 |
0 |
62 |
. |
| Nov 05 |
@ChrisEsplin I mostly agree, but this seemed like the kind of comment that warranted either a non-response or a calm "Ah she is being naive here" versus the vitriol and anger that many seemed to have.
|
1 |
0 |
0 |
0 |
200 |
. |
| Nov 05 |
Someone's doing the thing! I haven't actually tried @wabi yet but this is pretty much what I was talking about in 2021 https://t.co/XoGmNIhPI1
|
3 |
0 |
0 |
279 |
142 |
. |
| Nov 04 |
@maxjanderson Have you used the product? This is a great example of a “looks smart but is actually retarded” tweet
|
3 |
0 |
0 |
0 |
114 |
. |
| Nov 04 |
@jackfriks Agree 100%. He missed step 0: make something genuinely good - high quality design - brand/ui/ux, good feature breadth and depth. Something that users feel justified in shelling out for. Even if retention is bad at the beggining as you’re figuring out the right mechanics, people…
|
1 |
0 |
0 |
0 |
290 |
. |
| Nov 03 |
@rom1trs you have a link to that article?
|
1 |
0 |
0 |
0 |
41 |
. |
| Nov 03 |
@shl Do you not struggle with google docs lack of ability to organize the documents into something that more resembles a docs site than a collection of random files? We landed on Gitbook for evergreen stuff (still use g docs for one time docs). Agree 100% on slack though
|
3 |
0 |
0 |
0 |
271 |
. |
| Nov 02 |
This is still the best quick-website-builder I've used - really really good and not as well-known as it should be. You should check it out!
|
7 |
2 |
1 |
882 |
139 |
. |
| Nov 01 |
People have gotten way too sensitive on this stuff - clip was a great example of the *right* way to broach a subject like this if you’re going to. She didn’t say the gov should take it, didn’t imply anyone’s a bad person for having it, just said “give it away homies”
|
0 |
0 |
1 |
360 |
267 |
. |
| Nov 01 |
@ChrisEsplin Interesting. I find it faster and more control to just tab auto complete for the boilerplate, but in js land there does tend to be more of it
|
0 |
0 |
0 |
0 |
154 |
. |
| Nov 01 |
Man AI has gotten crazy good
|
0 |
0 |
0 |
349 |
28 |
. |
| Nov 01 |
For people building software like this, where does design fit in? There’s no way the LLMs are giving smooth ui/ux on the first pass, is there?
|
1 |
0 |
3 |
534 |
142 |
. |
| Nov 01 |
A lot of people talk about products as though there’s some kind of perfect-information marketplace where everyone can come and buy the one-version of particular idea, and you need to be first to make the idea to secure the shelf space. But that’s just not how it works at all
|
2 |
0 |
0 |
321 |
275 |
. |
| Nov 01 |
So it’s not just me
|
1 |
0 |
0 |
383 |
19 |
. |
| Nov 01 |
Lots of talk that Apple missed Ai. And Siri is a truly awful product. But they still own the OS of 1.5b mobile consumers and growing, and have several years to learn from others early consumer AI experiments & release something incredible. I think they probably will https://t.co/evqlyUwxgg
|
2 |
0 |
0 |
147 |
294 |
. |
| Oct 28 |
This matches my own general intuition that consciousness ain’t that special. Our brains got big enough to assign symbols to stuff, which was physical stuff first but then we started assigning words to non physical stuff like emotions and systems, and we stopped noticing the difference between words that represent real things and words that represent non-physical-things, and terms like “consciousness”, which is not attached to the physical world, became so vague as to assume a kind of magical quality.
At some point we learned how to use language inside our heads, and gained the belief that the voice we “hear” as we do that is our “self”. Combine that belief with a recursive loop of observing and narrating, and you have an emergent thing. That’s not to say human consciousness isn’t *different* to silicon consciousness, or that we should treat silicon consciousness with the same reverence that we do human, but it does feel like comparing flying - birds can do it or planes can do it, but it’s still flying.
|
3 |
0 |
0 |
388 |
1k |
. |
| Oct 28 |
@lottsnomad I thought that too and then @faith_alqah sent me a 100 screen Figma and we built it anyway and it turned out she was right (about this one) I think a big reason why people convert and it’s rated so well is because the app feels reasonably well finished and substantial, which I’m…
|
1 |
0 |
0 |
0 |
292 |
. |
| Oct 27 |
@JosephKChoi Yusss!
|
1 |
0 |
0 |
0 |
19 |
. |
| Oct 26 |
@karpathy re: your gym-vs-education analogy. Do you think people would still go to the gym at same rate if their progress wasn’t physically visible to them/others and/or if they didn’t feel better? Not sure it holds in that way
|
0 |
0 |
0 |
0 |
227 |
. |
| Oct 26 |
Still haven't found a way to reliably get good AI inside of slack. Claude's own implementation is so bad you'd wonder how anyone ok'd it for shipping. https://t.co/rH8CXuv6H0
|
0 |
0 |
0 |
192 |
174 |
. |
| Oct 26 |
Two ingredients to a coming explosion in actually-useful browser agent products:
1. The Browser Agents getting good enough
2. Building the UI/UX scaffolding around it
It sounds like hubris but I genuinely think we're furthest along on #2 of anyone I've seen. Just waiting for a breakthrough in #1 and I can feel it coming
|
3 |
0 |
1 |
281 |
323 |
. |
| Oct 26 |
Two ingredients to a coming explosion in actually-useful browser agent products:
1. The Browser Agents getting good enough
2. Building the UI/UX scaffolding around it
It sounds like hubris but I genuinely think we're furthest along on #2. Just waiting for a breakthrough in #1…
|
0 |
0 |
0 |
0 |
279 |
. |
| Oct 26 |
😂 what the hell, just bought an Anker today for 6x this
Sold out on Amazon for now 😞
|
0 |
0 |
0 |
450 |
86 |
. |
| Oct 23 |
The new pattern of measuring AI progress by how many minutes it can work consecutively for reminds me of… https://t.co/LYyM4iurpp
|
1 |
0 |
0 |
170 |
129 |
. |
| Oct 21 |
These guys have a ton of twitter hype and spent last week announcing new features on here, but their app literally doesn't work. Their support email basically told me "Works for us"
Sign of the times 🤡 https://t.co/jiuiN6fx6J
|
2 |
0 |
0 |
238 |
226 |
. |
| Oct 17 |
Another few hours trying new AI tools today - this time the AI avatar tools - every single one littered with bugs in the onboarding.
The enshittification continues... https://t.co/Faxp9CuN27
|
0 |
0 |
0 |
390 |
192 |
. |
| Oct 16 |
@mamagnus00 Accuracy
|
0 |
0 |
0 |
0 |
20 |
. |
| Oct 15 |
@kwharrison13 @pitdesi +1
|
1 |
0 |
0 |
0 |
25 |
. |
| Oct 15 |
Couldn't make this up - logged on this morning excited to try out some new AI data analyst tools - *both* of the ones I've heard of have critical bugs on the onboarding flow https://t.co/7bdtlXKXf3
|
1 |
0 |
0 |
897 |
197 |
. |
| Oct 15 |
Isn’t it kind of bad that the AI czar is openly conspiratorial against the second most important AI company in the world and the US? Shouldn’t he at the very least be engaging in dialogue?
|
1 |
0 |
0 |
412 |
188 |
. |
| Oct 13 |
@herreraevel Emailed you - assume it’s fully booked?
|
0 |
0 |
0 |
0 |
52 |
. |
| Oct 13 |
Ryan himself wondered about this back in May, and don’t think Remix 3 really disproved this hunch. If you’re on react and thinking “how can I improve the ratio of complexity:value”, there’s more value in going full offline first than in using a slightly better React, IMO https://t.co/oheqhHL0x2
|
1 |
0 |
0 |
409 |
295 |
. |
| Oct 13 |
I love that the remix guys are popularising web standards again among js folks, but I (currently) feel that the improvements they’ve made are in a dead zone. Marginally better than react, but not a step change in simplicity or functionality yet
|
2 |
0 |
0 |
849 |
244 |
. |
| Oct 13 |
Can someone tell me who all these tweets are subtweeting?
|
0 |
0 |
1 |
698 |
57 |
. |
| Oct 10 |
I'm slowly changing my mind on how much more value we can squeeze out of AI with the current architecture (LLMs).
Human inquiry/reasoning has multiple steps - we have a "hunch", then we get our thoughts out of our heads and verify them (e.g. writing an article or writing a maths proof). Then there's a loop as our reasoning modifies the original premise until it becomes logically coherent.
LLMs are already very good at the "hunch" phase. But we expect the base model to also do the verification step and come out fully formed. Surely it's better to just treat that phase separately and run it through the same loop *after* it comes out of the model?
|
2 |
0 |
0 |
467 |
654 |
. |
| Oct 10 |
https://t.co/CetlanOWOu
|
0 |
0 |
0 |
104 |
23 |
. |
| Oct 10 |
Another paper on HN today highlighting the shortcomings of Reasoning Models.
My synthesis (Ironically summarised with Claude):
- Humans use deductive reasoning with state tracking - when you rule something out, it stays ruled out in your working memory. Your mental model updates and enforces consistency - ruled-out causes become hard constraints on all future reasoning.
- LLMs use sequential probabilistic reasoning - they can rule something out but may "forget" and re-consider it later because they lack hard constraints.
- LLMs have soft probabilistic weighting instead of explicit working memory - earlier conclusions influence but don't guarantee exclusion.
- Both are "reasoning", just with different architectures and failure modes.
Can LLMs ever do perfect logical reasoning?
- Pure LLM architectures will likely always have limitations because they generate text token-by-token based on probabilities, not by executing logical operations
- They lack explicit data structures for tracking ruled-out hypotheses or maintaining symbolic state between tokens
|
0 |
0 |
1 |
665 |
1.1k |
. |
| Oct 07 |
If I keep recommending @magicpatterns to people maybe one day they'll build the feature that let's me use my own css-based design system 🤞 https://t.co/Pl3Cc12bCJ
|
2 |
0 |
1 |
193 |
162 |
. |
| Oct 03 |
Waiting for Teve and Ve Jobs to take over the timeline
|
1 |
0 |
0 |
207 |
54 |
. |
| Oct 03 |
@bradgessler @37signals 👋 not using Hotwire native, built our own on react native but using rails web views & sending html over the wire. One app with hundreds of 5 star ratings, another launching soon
|
7 |
0 |
0 |
0 |
205 |
. |
| Sep 30 |
"How do you manage 8 dev teams at the same time?"
Every morning: Headphones on, read updates, make loom responses for 30m, then jump on call with whoever needs a deeper dive.
+ Plus super talented team
+ Plus 3+ years of experiments
+ Plus clear guidelines on how to build https://t.co/Xy9LUG221Z
|
7 |
0 |
0 |
426 |
299 |
. |
| Sep 28 |
@lottsnomad 👋 growi I believe
|
1 |
0 |
0 |
0 |
29 |
. |
| Sep 26 |
Over time I've learned:
When you see controversy on the TL, it's not cowardly to take a deep breath, let it blow over, and contribute later when it's less heated, & where you won't accidentally find yourself contributing to the polarization
|
1 |
0 |
1 |
407 |
244 |
. |
| Sep 26 |
Every single time there's controversy on the timeline, I somehow only feel compelled to share the viewpoints of the side I mostly *don't* agree with, because the side I agree with is so over represented, finding a good point from the other side feels like it should be amplified
|
2 |
0 |
0 |
650 |
278 |
. |
| Sep 24 |
That's a lot of bookmarks https://t.co/LwAArYwEDA
|
0 |
0 |
0 |
239 |
49 |
. |
