building NoClutterCRM 17+ years shipping software, now learning to sell it. I talk about AI and Softwares.
Joined April 2022
- Tweets764
- Following339
- Followers253
- Likes508
tim peters died. he wrote the zen of python and timsort, which java and android ended up using too. on the python mailing list they called him timbot because he answered everyone so patiently. rip
1
44
a team lead wrote that their team stopped reviewing each other's prs. claude writes it, the person on the ticket reviews it and merges. their gut said bad idea, a month in its working fine.
solo devs have been shipping like that forever lol
5
2
163
united is banning ai agents from booking flights, says its a privacy risk. southwest went the other way and lets chatgpt users see its flights. i think southwest wins this one, people book where their assistant can actually do the booking
1
40
uk regulator wants universities back on written exams because of ai. one of the other fixes they listed is grading students on how well they edit ai output. interesting.
1
1
48
each founder sells about $250M and keeps around $18B. thats under 2% of what they own. if i built claude i'd sell a little too lol
I believe this is the largest founder sale ever in an IPO in history ?
$ANT
And I do NOT believe the sales are to cover taxes (Zuck's were).
And I believe that benevolent founders did not allow their employees to sell the same percentage.
And I do believe that employees can sell 10% of their shares on Day 1.
--
why would anyone about to cure all diseases and unleash a "decade of miracles" want or need to sell ~$1.75Bn worth of stock onto retail for ?
52
The Center for AI Safety just updated the Remote Labor Index. Fable 5.2 completed 41.3% of real freelance projects at a level a paying client would accept. A year ago the best agent managed 2.5%. Opus 5.5 is at about 21%.
That jump is real, I am not going to downplay it.
The detail that interesting is how they still grade the work. They tried letting another AI check the submissions, and it kept getting fooled by work that looked finished but was actually broken. So humans still open the files in the real software and compare them with what a freelancer delivered.
I keep running into a smaller version of this. The summary says the work is done, and then you open it and something is off. Looking finished and being usable are still two different things.
When we launched the Remote Labor Index a year ago with @ScaleAILabs , the best AI could fully automate 2.5% of real freelance projects. Fable 5.2 now automates 41.3%.
The automation rate is doubling every ~3 months. AIs may soon be able to do all remote work faster and more cheaply than humans.
1
43
the typing part of coding is almost gone.
the "is this actually right" part is now most of my day.
16
ai can write 200 tests in a minute and every single one passes.
then a real user clicks something nobody thought of and it all falls apart.
green coverage isn't the same as working software.
read a post on hn about ai making codebases fragile. the shallow tests part is the real one for me. ai writes tests for getters and setters, coverage looks great, and the stuff users actually do nobody tested
1
18
chrome, edge, firefox and mdn teams got together and made a skill so coding agents stop writing old javascript. their reason is the training data is mostly legacy code. i mean the browser teams are writing docs for the agent now, not for us lol
1
26
said this morning open models would hit opus 5.5 level in 3-4 months. meituan just dropped longcat 2.5, open weights, MIT license. close, but they compared it to opus 5 not 5.5. clock is still running.
We’re releasing the weights for LongCat-2.5 today.
· 1.6T total parameters, ~48B active, natively multimodal
· Stronger visual understanding and agentic coding in one model
· From complex coding projects to creative design and 3D tasks
Released under MIT. Weights are on Hugging Face: huggingface.co/meituan-longc…
36
The AI war isn’t about intelligence anymore. Opus 5.5 already does almost everything. Now it’s about token efficiency and price.
The real liberation comes when open models hit this level, with no silent nerfs and dirt-cheap inference. My bet: 3–4 months.
2
1
159
I don't feel same excited for it anymore and reason is Opus 5.5 it is so good and so cheap comparatively that it gets done everything I need and still the usage is left for week.
i may use fable-5.2 for planning for some things but (not really need to) opus-5.5 is daily driver as has been since it came out. i just want they not keep nerfing it.
Introducing Claude Fable 5.2 and Claude Mythos 5.2, designed to tackle the most daunting technical challenges.
They’re capable collaborators on ambitious projects in many fields, and show improvements in areas like coding, reasoning, and vision.
1
51
millions of users and product hunt still got down to nine months of runway. man building a business is hard.
In 2016, Product Hunt almost died.
Here's what happened and what I learned. Hope it's useful for other founders:
ryanhoover.me/post/how-produ…
2
77
elon says grok bot is getting fixes. i mean good, but he himself said the limits are because it costs too much to serve. i dont think thats getting fixed overnight.
Just did a round table review with the Grok @Bot engineering team.
Fixes are being implemented overnight and you should notice a dramatic improvement by tomorrow. If not, reply to this post and we will address any remaining issues right away.
36