Paras

182 posts

Paras

Paras

@mumtor99

Perpetual Builder. Prev: Amazon building Ads. Qualcomm building CPUs. https://t.co/dZ1Nn3SCrO

Mumbai Katılım Eylül 2024
95 Takip Edilen5 Takipçiler
Paras
Paras@mumtor99·
And on the other side, Claude increasingly feels like the opposite. The models feel less impressive than they should. Claude Code / desktop / newer surfaces often feel brittle, under-tested, and weirdly low-taste for products being shipped at this scale. OpenAI feels like a company trying to build a lasting AI platform. Anthropic increasingly feels like a company being pulled by enterprise monetization and IPO gravity.
English
0
0
0
13
Paras
Paras@mumtor99·
That’s why OpenAI feels stronger to me right now. Not just because of the models — GPT-5.4 clearly helped — but because the products feel like they’re being pushed by real user judgment. Codex / Codex App is a good example. It’s no longer just a narrow coding tool. It’s starting to feel like a real super app for builders.
English
1
0
0
9
Paras
Paras@mumtor99·
My current read on Anthropic: they’re falling into the classic enterprise-company trap. When your biggest wins come from large org budgets, the feedback loop changes. The buyer is not the same as the user. And the user is usually not paying with the same level of scrutiny as a person spending $20 or $200 of their own money.
English
1
0
0
6
Paras
Paras@mumtor99·
That’s genuinely commendable. Building something means enduring constant ups and downs, and you’ve done an impressive job staying with it. I see Repo Prompt as one of Canada’s original AI startups, even if most people don’t fully appreciate that. It may not be obvious, but I’m Canadian by naturalization and try to keep up with the scene there. You’re always near the top of that list.
English
1
0
1
150
eric provencher
eric provencher@pvncher·
Been grinding on @RepoPrompt for nearly 2 years now. Full of ups and downs. I just shipped the orchestration workflow, and it feels like it's finally come together where out of the box it can deliver a magical experience. Really proud of where it is now, and I hope you try it!
English
15
5
93
2K
Paras
Paras@mumtor99·
I am deeply concerned by Ray Kurzweil’s recent claim that, over time, people will come to believe AI models are conscious. What saddens me most is how easily many seem to overlook something fundamental: consciousness is a quality of living beings, and there is something profoundly important about that distinction. We may be entering a troubling period, not just for the next few decades but perhaps for centuries, in which many people will treat AI as if it were conscious or alive when it is neither. I find that deeply unsettling, and I worry about the cultural and moral consequences of that confusion. As for the hope that AI will soon bring widespread abundance, I believe that expectation is far too optimistic. True abundance, if it comes at all, is likely still centuries away. Those who insist it will arrive within the next few years are either deeply mistaken or misleading others.
English
0
0
0
10
Paras retweetledi
eric provencher
eric provencher@pvncher·
Opus 4.7 is the first claude model I've seen make incorrect mcp tool calls.
English
27
23
787
87.3K
Paras
Paras@mumtor99·
@FactoryAI Congratulations! Well deserved and an awesome follow on round! I would love to use Droid more than my Pro Plan allowance. Could you entice me with a promo code for the Max Plan?
English
0
0
4
1.3K
Factory
Factory@FactoryAI·
Today, we are excited to announce our $150M Series C led by Khosla Ventures with strong participation from Sequoia Capital, Blackstone, Insight Partners, Evantic Capital, Abstract Ventures, 20VC, NEA, and Mantis VC. This puts our valuation at $1.5B and will accelerate our investment in research, product, and global go-to-market. Long live developers.
Matan Grinberg@matanSF

x.com/i/article/2044…

English
49
39
616
210K
Paras
Paras@mumtor99·
Just watched @dwarkesh_sp’s interview with Jensen Huang. It felt like Dwarkesh was carrying water for Anthropic, and Jensen spotted it almost immediately. Most of Jensen’s pushback seemed aimed at Dario Amodei’s rhetoric. Pretty disappointing to watch.
English
0
0
0
38
Paras
Paras@mumtor99·
Saint @thsottiaux, the dark side is tempting me with their new shiny thing. when will you shine the new light upon us? save me!
English
0
0
0
16
Paras
Paras@mumtor99·
@pvncher didn't they claim it's better at instruction following and may need a prompt tune up? but seems like you are saying it's not good at instruction following. what's happening at anthropic!
English
1
0
2
3.7K
eric provencher
eric provencher@pvncher·
First impression of Opus 4.7 is that it doesn't give a crap about your workflow prompts and just does what it wants
English
49
12
540
37.7K
Paras
Paras@mumtor99·
@benvargas are there any changes happening on the auto compact on the compact end point?
English
0
0
0
48
Ben Vargas
Ben Vargas@benvargas·
The OpenAI/Codex team has done so well on compaction, I almost forget the days of worrying about compaction. This was ServiceNow work, not super hard or crazy amounts of coding, but it didn't get dumb, hallucinate, or go off the rails... just worked really well all day.
Ben Vargas@benvargas

@BennettBuhner @opencode I mean, it's not super crazy... I just closed a Codex CLI session at...

English
2
1
43
4K
Bindu Reddy
Bindu Reddy@bindureddy·
Super excited for a launch of an upcoming large LLM It BRILLIANTLY combines efficiency and performance on agentic loops People won't stop talking about it 🤯
English
34
12
255
17.9K
Paras
Paras@mumtor99·
We have a good way to start long-running background tasks, but we still need a good way to wait for them. Today, monitoring is too token-heavy: agents poll, waste tokens, and sometimes stop checking. The better pattern is event-driven: wake the main agent only when the task emits a specific signal. Cursor added something like this recently, and it’s exactly what tools like Codex need for jobs that run for hours.
English
1
0
3
398
Thomas Ricouard
Thomas Ricouard@Dimillian·
I’m diligently working on enhancing all the features of the Codex app to make it even better for you. If you have any suggestions or ideas for improvement, please feel free to share them with us here!
English
432
11
637
50.5K
Paras
Paras@mumtor99·
@jsnnsa omniscience would be alien to humans
English
0
0
1
29
jacob
jacob@jsnnsa·
it's so unfortunate that pretraining gives such a strong prior over human-shaped reasoning globally optimal reasoning is probably quite alien!
English
1
0
1
404
Paras
Paras@mumtor99·
@tszzl misunderstood. big time.
English
0
0
0
5
Paras
Paras@mumtor99·
I’m honestly amazed by how much time GPT-5.4 Pro saves me, even when it takes an hour to finish a task. I use it to plan and review complex code changes, and it absolutely crushes it.
English
0
0
0
39
Paras
Paras@mumtor99·
@benvargas agreed that it feels a bit strange, but at least its original among the frontier subs. i think their reasoning must be something like "we are already so generous with limits, we don't need to do any gimmicks"
English
1
0
1
24
Ben Vargas
Ben Vargas@benvargas·
as a consumer... I'm more "comfortable" or "happy" knowing I paid Anthropic $200/mo even if $100 Max would have sufficed, because some months I might have higher needs and in those months I get 4x being on 200 Max. OpenAI there's no benefit, its just linear lock-in spend
English
1
0
5
447
Ben Vargas
Ben Vargas@benvargas·
Appreciate the clarification from OpenAI. In my mind, $100 is the sweet spot... on $200 you're committing to twice as much spend with no multiplier benefit. Anthropic $200 @ 20x vs $100 @ 5x, OAI is not sweetening the deal like that for their extra $100/mo use it or lose it.
Tibo@thsottiaux

Hi! Getting a lot of questions on Pro plan rate limits, so I wanted to clarify a few things: - Your Pro $100 plan includes at least 10x Plus usage, till May 31 with the 2X usage boost. - Your Pro $200 plan includes at least 20x Plus usage, till May 31 with the 2X usage boost. This is also the SAME usage this plan had since the 2x promo in February; we previously didn't document this explicitly. Now to address the confusion. Our pricing page says "5x or 20x usage", and many of you very understandably took it to mean "10x and 40x" given then 2X usage boost. That's our fault. We mixed up two things in a confusing way on that page - (1) sharing that the Pro $100 launched with a 2x boost, making it 10x Plus till May 31 and (2) sharing that Pro $200 retained its 2x boost and sharing for the first time that this equates to 20x Plus. We're going to update our pricing pages to make this clearer. Sorry it wasn't as clear as it should have in the first place.🫥

English
7
0
19
4.3K

Keşfet

@RepoPrompt @FactoryAI @dwarkesh_sp @thsottiaux @pvncher @benvargas @bindureddy @Dimillian