| May 21 |
Ok this is impressive from @cursor_ai using O3 Max.
We've been building a hybrid boilerplate (similar to hotwire native) and doing some weird atypical stuff.
Kept getting stuck on things that should be straightforward but weren't, and up to now Cursor was useless - none of the models seemed to *understand* what was happening.
Just tried with O3 Max (thanks @Shpigford) and it got me a working solution in one pass. Not only that, it managed to build an understanding from the code that the previous models couldn't.
|
10 |
1 |
1 |
2k |
522 |
. |
| May 21 |
Shopify announced their official Web Components today. I thought inline event handlers were frowned upon but I'm loving the locality of behaviour and Platform First approach here
https://t.co/DriTf8HLGd https://t.co/uRiId39pqy
|
0 |
0 |
1 |
222 |
227 |
. |
| May 21 |
I feel like the bar for customer service in most companies is "let's address issues had by large numbers of customers and dismiss outliers as complainers", when it should be "let's address every *reasonable* issue raised, including outliers"
|
1 |
0 |
0 |
137 |
241 |
. |
| May 21 |
Used @theblueground recently.
The Good: Their email response times are fast
The Bad:
Apt was unsleep-able for the first 5 days due to blinds that didn't work.
Trying to charge us a sneaky additional $260 fee for "2 hours of additional cleaning" despite leaving it exactly as described in their move out email.
|
0 |
0 |
0 |
160 |
311 |
. |
| May 21 |
I think @staysaasy has the highest amount of tweets-I-wish-id-written of any account. Simple insights clearly articulated. This one is 100% my experience
|
1 |
0 |
0 |
183 |
153 |
. |
| May 21 |
I absolutely loathe bureaucracy and regulation, but I share the view that thinking that’s what’s behind the EU/US divide is incorrect.
In the US, there is a pervasive desire for more-and-better, coupled with the absence of shame and cynicism (and then the obvious head start/rich-get-richer, network effects that compound that).
It’s not the one-true-way - both have benefits and trade offs, and I like the European way. But you don’t get US level wealth by simply removing over regulation and keeping the culture the same.
|
4 |
0 |
0 |
427 |
525 |
. |
| May 21 |
@Una @rachelandrew That's awesome! Any chance those slides are online anywhere?
|
0 |
0 |
0 |
0 |
79 |
. |
| May 20 |
@wagslane @thepatwalls would contest this I’m sure
|
2 |
0 |
0 |
0 |
50 |
. |
| May 20 |
@jackfriks It's crazy the 37 Signals material isn't more well known. @Wildbit was another one, but we need more. I'm trying too 😁
|
1 |
0 |
0 |
0 |
129 |
. |
| May 20 |
@therealdecross Depends but if I'm 5 nested files deep then yeah, but that seems more dependent on short term memory which for me is quite bad. But reading text feels like it shouldn't have the same issue. https://t.co/iuibk3LLv3
|
0 |
0 |
0 |
0 |
229 |
. |
| May 20 |
Am I the only one who finds it quite difficult to switch "brain modes" from coding or something else that's rapid-fire, to reading a long message or email and properly comprehending it? Is that a thing?
|
3 |
0 |
1 |
304 |
202 |
. |
| May 19 |
@derkolstad This is extremely offensive to my people. Do better
|
0 |
0 |
0 |
0 |
63 |
. |
| May 19 |
In Charlotte, North Carolina and never in my life have I seen as many tucked-in-collared-tees. I feel out of place not wearing one.
|
1 |
0 |
1 |
228 |
131 |
. |
| May 19 |
@bentossell Factory eyyyyy
|
2 |
0 |
0 |
0 |
26 |
. |
| May 19 |
@usescoutai @RobFarlow Study here: https://t.co/oSdmL7eiIm
|
0 |
0 |
0 |
112 |
58 |
. |
| May 19 |
In 2025, the top open source browser agent can still only reliably complete tasks 30% of the time. To replace real employees, this will need to get closer to 100%. The top closed-source is still only at 60%. This reflects our real world experiments at @usescoutai
HT @RobFarlow https://t.co/8T6Yuh5KOM
|
1 |
0 |
1 |
227 |
302 |
. |
| May 19 |
@RobFarlow Got you thanks. I've seen a few people imply vision based browser agents are better - but I can't find any/many examples - what ones were you referring to in your original tweet?
|
0 |
0 |
0 |
0 |
189 |
. |
| May 19 |
As AI agents get more common, there will come a day when we choose products based on how annoying their captcha is https://t.co/pBFMUtFyF3
|
1 |
0 |
0 |
144 |
138 |
. |
| May 19 |
@nlpnyc @lessin Do you not think that when they’re making toddler-level mistakes that the idea they’re “intelligent” (as we’d currently define intelligence) seems inaccurate? And therefore it’s reasonable to have some skepticism they’ll go from not-intelligent to intelligent?
|
0 |
0 |
0 |
0 |
276 |
. |
| May 19 |
@RobFarlow What are the better open source alternatives currently?
|
0 |
0 |
0 |
0 |
66 |
. |
| May 18 |
Took me a few watches but if you zoom in there are people on the masts/sails that get thrown off when it hits.
|
0 |
0 |
0 |
305 |
110 |
. |
| May 17 |
@joshmanders As much as this triggers me it is exceptionally well done and I commend whoever built it
|
3 |
0 |
0 |
0 |
101 |
. |
| May 14 |
RT @tonyennis: The history of web development:
1. Frameworks and tooling (abstractions) are built to do what the platform can't.
2. The pl…
|
0 |
5 |
0 |
0 |
140 |
. |
| May 14 |
Holy shit it’s happening https://t.co/OEzgQiISzV
|
4 |
0 |
1 |
462 |
48 |
. |
| May 14 |
Claude 3.5-Sonnet was released ~a year ago & despite being followed by 3.5-Haiku & 3.7-Sonnet, is *still* the recommended model for half of cursor tasks
- Incredibly impressive from the @AnthropicAI team
- Proof that we're perhaps not progressing at exponential speed toward AGI https://t.co/I8vAkIPugK
|
3 |
1 |
0 |
276 |
311 |
. |
| May 11 |
US prices in a nutshell https://t.co/85uTP1Ri2m
|
1 |
0 |
0 |
190 |
47 |
. |
| May 11 |
RT @peteromallet: I spent the past months building https://t.co/LNfjaMBCem.
Growing from the Banodoco community, my hope is that it’ll bec…
|
0 |
30 |
0 |
0 |
135 |
. |
| May 11 |
@devhenryhale @coderhq Haven’t seen lake.js - what drove that decision?
|
0 |
0 |
0 |
0 |
71 |
. |
| May 11 |
@devhenryhale @coderhq Adding to list to try over the weekend
|
1 |
0 |
0 |
0 |
61 |
. |
| May 10 |
Excited to announce I'll be speaking at Big Sky Dev Con in August 😊
Last year's speakers were some of the people I admire most in the industry, & I'm sure this year will be just as epic - still can't believe they let me in tbh, but I'll try not to disappoint
Come Join!
https://t.co/FJeFAmcVeX
|
7 |
0 |
0 |
376 |
298 |
. |
| May 10 |
@tinybirdco Maybe a silly question, but do you think there's a theoretical way to bring performance up to closer to 100, by optimizing the system prompt?
|
1 |
0 |
0 |
0 |
153 |
. |
| May 10 |
@Shpigford Check your DMs!
|
0 |
0 |
0 |
0 |
26 |
. |
| May 10 |
@shl Congrats man! ❤️
|
1 |
0 |
0 |
0 |
21 |
. |
| May 09 |
@bentossell you should've stayed longer!
|
0 |
0 |
1 |
163 |
40 |
. |
| May 09 |
Twitter & San Francisco this week are basically the same thing. Keep meeting people at events then seeing them on my feed - what I imagine early twitter must have felt like.
|
4 |
0 |
1 |
319 |
177 |
. |
| May 09 |
@shreyashdamania @codegen It's in the screenshot :)
|
1 |
0 |
0 |
0 |
51 |
. |
| May 08 |
Glorifying drinking is cringe
Glorifying never-drinking is cringe
Normalize nuance
Denounce dichotomozation
|
2 |
0 |
1 |
472 |
109 |
. |
| May 08 |
@coreyhainesco @shl Much of the original reasoning came from the fact that - for LLMs - typescript is more robust (if it compiles that usually means it works) has much more training data so can one-shot things quicker, has better UI libraries, and better DX (autocomplete etc). I showed him that with…
|
8 |
0 |
0 |
0 |
301 |
. |
| May 08 |
The arc is complete. Kudos to @shl for sticking to his promise and keeping an open mind. Turns out you can change people’s minds on the internet! 😊 https://t.co/cq2U8ULNct
|
112 |
4 |
4 |
14.3k |
171 |
. |
| May 08 |
If this is true, why do the models still not seem to understand reality?
|
0 |
0 |
0 |
342 |
72 |
. |
| May 08 |
😂 some people can't help themselves once they catch the faintest scent of Schadenfreude. The 37signals owners have more wealth (measured in dollars and freedom) than 90% of founders, which is exactly why they indulge in projects like this. They’ll be just fine 😂
|
7 |
0 |
1 |
849 |
262 |
. |
| May 08 |
Blasphemy!
|
0 |
0 |
0 |
200 |
10 |
. |
| May 07 |
I've gotten so tired of this that yesterday I built a custom UI for my bank data so that I can see it without logging in, and we'll likely migrate off @Xero soon because it takes on avg 1 minute to simply get to the home screen
|
0 |
0 |
0 |
223 |
227 |
. |
| May 07 |
@kingsley_kelly The point was I wouldn't have intuitively guessed that *entering text into a plain text field* was the part that doesn't work reliably
|
0 |
0 |
0 |
0 |
150 |
. |
| May 07 |
This seems more correct to me than “models have human-like intelligence”
|
4 |
0 |
0 |
260 |
72 |
. |
| May 07 |
I try not to be mean on here, but it’s hard to highlight obviously wrong statements like “Basically everything you see will be generated by an Llm” without coming off as that. That idea - that humans don’t care about predictability in their interfaces - defies a *lot* of patterns to date. If you think it, write a post or tweet thread that breaks down why - these offhand comments making big claims come off as extremely hubristic and naive.
|
0 |
0 |
0 |
272 |
442 |
. |
| May 07 |
The Promise...
The Reality...
Browser Use agents are still *extremely* rough https://t.co/Z3R1E9lvbJ
|
1 |
0 |
1 |
189 |
102 |
. |
| May 07 |
😂 The @AnthropicAI system prompt basically does what I did with https://t.co/NESs4jkTw4 https://t.co/IHSo4EN499
|
1 |
0 |
0 |
207 |
111 |
. |
| May 06 |
Wait what? @windsurf_ai is the html first ai ide? Must go back and use it more
|
1 |
0 |
1 |
286 |
78 |
. |
| May 06 |
https://t.co/N4qcduSVia
|
1 |
0 |
0 |
127 |
23 |
. |
| May 06 |
Is @OpenAI neutering the models available through the API? Asked o3 a really tricky code reasoning problem in the chat app - correctly identified and fixed everything. Used the exact same prompt in @cursor_ai with O3 and it totally missed it and kept going around in circles
|
0 |
0 |
0 |
220 |
274 |
. |
| May 05 |
@searchbrat @levelsio As an Irishman who did gtfo - Amsterdam is 100% the place. Similar prices to Ireland but 10x quality of living. Pieter may disagree - not sure he’s a fan of NL. We did a euro tour before deciding
|
4 |
0 |
0 |
0 |
217 |
. |
| May 04 |
@rauldoesnothing Also I think it's open access now?
|
0 |
0 |
0 |
0 |
51 |
. |
| May 04 |
@rauldoesnothing Have access. Was impressed with the onboarding (they have you set up a cloud VM which is the feedback-loop-closer), but then the first task I assigned the agent - it didn't check it's work in the browser and produced code that didn't work
|
0 |
0 |
0 |
0 |
255 |
. |
| May 02 |
@nickbakeddesign Pretty sure there was a 6713 in there at one point? (because I copied it)
|
1 |
0 |
0 |
0 |
90 |
. |
| May 02 |
@JosephKChoi @consumerclub_ @lottsnomad 😂 you must’ve been the one person I didn’t chat to. Thanks for hosting though!
|
2 |
0 |
0 |
0 |
118 |
. |
| May 02 |
Whatever LLMs do is not *reasoning*. Human intelligence doesn't work like this. If a human said this you would think they were having a stroke. https://t.co/0jL8wSZkHb
|
0 |
0 |
0 |
202 |
167 |
. |
| May 02 |
@realcalebwin Looks very cool! Have you run any benchmarks (e.g WebVoyager) against it?
|
0 |
0 |
0 |
0 |
87 |
. |
| Apr 30 |
Honest question - if the box disables all the drones, what is capturing the aerial footage?
|
1 |
0 |
1 |
291 |
91 |
. |
| Apr 29 |
@adbc_an @codeiumdev @cursor_ai @cline Couldn't you say that about MCP?
|
0 |
0 |
0 |
0 |
71 |
. |
| Apr 29 |
This is exceptionally dumb. @codeiumdev @cursor_ai @cline will you guys figure this out? First mover gets to define the standard a la MCP https://t.co/yGgYbyw3Gj
|
6 |
0 |
7 |
673 |
161 |
. |
| Apr 29 |
@nikitabier @believeapp (Big fan of your work, and @solana)
As an outsider, this mechanism (random's respond to a tweet without the founder's buy-in, with a $ symbol), no url, no website, looks like crypto/memecoin spam - might that not blunt the "will continue to spread virally until saturated"…
|
2 |
0 |
0 |
0 |
298 |
. |
| Apr 29 |
@shl @ofcboogeyman It was originally an internal plan though right? Assuming that changed, would love to hear about what changed your guys' mind - looking at the sheer size of the codebase, assuming it's just a priorities thing? https://t.co/DPicKn2zNG
|
0 |
0 |
0 |
0 |
252 |
. |
| Apr 29 |
@shl @ofcboogeyman @shl are you guys still planning on migrating the gumroad backend to Typescript from Rails?
|
0 |
0 |
0 |
0 |
110 |
. |
| Apr 28 |
@TylerAlterman @Shadow_Rebbe Another (very different) one for you: https://t.co/xlYgVfUYma
|
1 |
0 |
0 |
0 |
90 |
. |
| Apr 27 |
@elidourado Are you using it a lot? Fancy trying an alternative that has better reliability?
|
0 |
0 |
0 |
0 |
92 |
. |
| Apr 27 |
😂 these lawyers are likely being paid 8 figures. Crazy
|
1 |
0 |
0 |
302 |
54 |
. |
| Apr 26 |
This is next level trust building and more people should do it. The lack of humility and honesty about shortcomings with all the new AI products is exhausting and leads to constant disappointment. The best AI powered products aren’t the ones that everyone’s talking about
|
1 |
0 |
0 |
217 |
271 |
. |
| Apr 24 |
@AirbnbHelp As expected, the DM thing is an exercise to save face, not to actually help https://t.co/Ms4Pq6TPGX
|
0 |
0 |
0 |
0 |
111 |
. |
| Apr 24 |
@AirbnbHelp I already spoke to support over the phone. They told me I hat to wait 30 minutes til they reach out to the host, and I’d get 20% off a re book which was no good because there was nothing else available, just used https://t.co/b9687JOAoQ
|
0 |
0 |
0 |
0 |
248 |
. |
| Apr 24 |
Had my first scam experience on @Airbnb today. Arrived to hotel and they had no record of my booking. Some dude just listed the place and took my cash it seems. Didn’t know that could happen on Airbnb but today I learned
|
1 |
0 |
5 |
382 |
220 |
. |
| Apr 24 |
@bentossell That tweet’s targeted at people who write code but end up creating long term unmaintainable codebases because they add libraries and complexity that compounds, which they wouldn’t if they understood what they were doing
|
0 |
0 |
0 |
0 |
231 |
. |
| Apr 24 |
Amen 👇
|
0 |
0 |
0 |
206 |
6 |
. |
| Apr 23 |
@adamwathan Not sure I follow
- Button in form - visually obvious I'm in a form because I've already input form data
- <a>s that looks like a button - I can cmd+click
What are other scenarios where a form button would appear that could be mistaken for a link button?
|
2 |
0 |
0 |
0 |
275 |
. |
| Apr 23 |
YES. A lot of the newer libraries and the hip products (cough @linear) use cursor: default and it feels off.
|
0 |
0 |
0 |
224 |
108 |
. |
| Apr 23 |
Thing I've used @usescoutai to do in the last 3 days
- Send multiple emails ✅
- Send twitter DMs ✅
- Send whatsapps ✅
- Check my Linkedin DMs and send a calendar invite ✅
- Attempt to book Uber eats (got stuck)
Slowly iterating through the brittle parts but very excited
|
3 |
0 |
0 |
153 |
271 |
. |
| Apr 22 |
@AlexKlarfeld Love what you're doing - new site looks great!
|
1 |
0 |
0 |
0 |
60 |
. |
| Apr 22 |
Made a thread 4 yrs ago w/ products I liked which had stackblitz/@boltdotnew (rocketship), @replit ($3b company), & @supabase ($2b company). All small at the time.
Good investmt thesis in hindsight: Technical founders with taste using hard tech to make complex things simple.
|
2 |
0 |
0 |
284 |
280 |
. |
| Apr 22 |
@CiaranHan is the man
https://t.co/viM9CFv7TR
|
1 |
0 |
0 |
0 |
46 |
. |
| Apr 20 |
"How to build an agent" (https://t.co/H98pbsdN7S) + iterating to get this prompt + high velocity eng team + really good UX owner/s who are also technical + good GTM with a few strokes of luck = potential $3,000,000,000 acquisition. Easy to think "there's no magic tech" but actually the chances of getting all these things right is extremely rare
|
1 |
0 |
0 |
368 |
346 |
. |
| Apr 20 |
Tried O3 ChatGPT - it's really good
What it shows is that we don't need super intelligence to have super utility - it's genius is in how well it can gather, summarise, and present data, a.k.a how seamlessly the model, the UI, and tool use are integrated.
|
0 |
0 |
0 |
226 |
255 |
. |
| Apr 19 |
Every now and then I try to find a password manager I don't hate, and every time I'm disappointed. Tempted to build one - requirements:
- Don't be slow
- Don't ask me to keep logging in with no way to dial down
- Use the chrome sidebar API so I don't lose what I was doing.
|
0 |
0 |
0 |
204 |
273 |
. |
| Apr 19 |
@GergelyOrosz This has always been a thing, for at least the last 15 years - I had a friend get sent home from the border on returning from a road trip to Canada, and barred for 10 years. If you’re on an Esta, work is an absolute no-go and always has been. Seems like there’s a lot of…
|
8 |
0 |
0 |
0 |
285 |
. |
| Apr 17 |
By far the most promising tool so far has been @codegen - integrates with Slack, lets you tag it, supports threading and follow ups. The out-of-the-box code reviews were much higher quality than @coderabbitai
Major gap is it doesn't support multiple github orgs through Slack
|
1 |
0 |
1 |
242 |
277 |
. |
| Apr 17 |
Have been experimenting with more AI dev tools recently
@DevinAI: Setup was extremely promising - they had me provision a cloud VM & install the packages it needs, but the first task I gave it - it didn't actually test the code changes against the live app & the PR was bad 🤷♂️
|
0 |
0 |
0 |
399 |
288 |
. |
| Apr 17 |
@AravSrinivas Question: Why build this as a new browser vs a cloud browser that someone opens once & logs into their services, then interacts with via tasks/chat to ask it to do stuff?
|
0 |
0 |
0 |
0 |
188 |
. |
| Apr 17 |
@Jthysamhume3 Hey! This still available?
|
0 |
0 |
0 |
0 |
40 |
. |
| Apr 17 |
A big part of what's allowed us to scale @rlygoodsoftware is settling on tools and sticking to them - @heroku is still our go-to, we're amassing quite a collection. https://t.co/VSFFgwMDvA
|
2 |
1 |
1 |
266 |
188 |
. |
| Apr 16 |
I'm looking to purchase a last minute @ReactMiamiConf ticket, potentially from someone who has one but can't make it. DMs open 😁
|
1 |
0 |
0 |
184 |
128 |
. |
| Apr 16 |
@kraustifer @ReactMiamiConf Hey! Are you selling?
|
0 |
0 |
0 |
0 |
49 |
. |
| Apr 15 |
Wild
|
0 |
0 |
0 |
208 |
4 |
. |
| Apr 14 |
@scrimba Any plans for ruby?
|
1 |
0 |
0 |
0 |
28 |
. |
| Apr 14 |
@patjfin @ycombinator Not to burst your bubble but did the same in Amsterdam, sans Waymo 😬
|
0 |
0 |
0 |
0 |
90 |
. |
| Apr 13 |
True. Now apply this to software engineering
|
0 |
0 |
0 |
264 |
44 |
. |
| Apr 11 |
@derkolstad We will be here! Give us a shout for sure!
|
1 |
0 |
0 |
0 |
54 |
. |
| Apr 11 |
@marcoroth_ @rubykaigi @railsconf Oooh please share some sneak peaks this sounds suuuper interesting
|
2 |
0 |
0 |
0 |
100 |
. |
| Apr 10 |
Who do I know in SF that's down to hang out this week or next?
|
1 |
0 |
1 |
224 |
62 |
. |
| Apr 10 |
Every time I think about this tweet I cringe. I was so wrong about this. Azure has a slightly nicer UI than AWS but the cloud platforms are so user-hostile it's crazy https://t.co/fBshRhWJoQ
|
0 |
0 |
1 |
271 |
190 |
. |
| Apr 10 |
Lots of talk of High Agency on the timeline - this is a very good example of it in action
|
1 |
0 |
0 |
181 |
89 |
. |
| Apr 10 |
@peteromallet Proud of you
|
1 |
0 |
0 |
0 |
26 |
. |
