Vijay

3.8K posts

Vijay banner
Vijay

Vijay

@imvijaymode

Building : @chainshorts | One Rule: Hunt or Be Hunted

San Francisco, CA Katılım Mayıs 2016
5.5K Takip Edilen646 Takipçiler
Vijay
Vijay@imvijaymode·
Open sourcing the weights is marketing. It gets developers excited, builds hype, and gets the model into tools and benchmarks. That draws more people to their paid API, which is where they make money. Their revenue comes from people using the hosted version, not from the open weights. They already hit three hundred million dollars in annual recurring revenue from API demand. When they open source the weights, that's usually a one-time snapshot. Future improvements, bug fixes, new features, and retraining on fresh data? Those come through their hosted API first, and stay under their control. Users of the API get the latest and greatest without lifting a finger Hosting a open-weight model on your own is too much expensive . So they just choose the existing Kimi subscription or Hosted API from an official provider.
English
1
0
2
122
Connor
Connor@ConnorTalksAI·
@dee_bosa I genuinely don’t understand how they will make money after they open-source it, right? wouldn’t people run it with their own computers or find a cheaper provider than buying from Kimi?
English
6
0
6
1.8K
Deirdre Bosa
Deirdre Bosa@dee_bosa·
And they said you couldn’t build a business on open source
Kimi.ai@Kimi_Moonshot

Kimi K3 has received far more love than we expected, and our GPUs are feeling it. Over the past 48 hours, demand has pushed close to the limits of our current capacity. To protect the experience of existing subscribers, we're temporarily pausing new subscriptions and prioritizing compute for current members. Existing subscribed users are not affected. We're adding capacity as fast as we can and will reopen new subscription spots in batches. Going forward, we'll also split membership into two more focused plans: Kimi Membership for Kimi Web, App, and Work; and Kimi Code Membership for coding workflows. This will help us match compute more precisely and keep the experience stable. Thank you for your patience and understanding!

English
52
35
514
109.3K
Udit Goenka
Udit Goenka@iuditg·
I absolutely love Codex but the only issue I see with Codex 5.6 Sol model is the token consumption is insanely crazy. Once the reset stops, it will be tough to continue using it unless Codex team continues to optimize the inference.
English
49
9
421
28.4K
Vijay
Vijay@imvijaymode·
@balajis Who is she. Never saw her before
English
0
0
0
109
Balaji
Balaji@balajis·
India is building startup cities.
Balaji tweet media
English
765
569
8.6K
1.6M
Claude
Claude@claudeai·
The Built with Claude: Life Sciences hackathon comes to an end. Thank you to everyone who spent a week building with Claude Science and Claude Code, and to @gladstoneinst and @cerebral_valley for co-hosting. Introducing the winners:
English
291
159
2K
416K
Vijay
Vijay@imvijaymode·
@sorower01 @AnthropicAI Getting same message . From settings >> usage - fable 5 sessions limits were not showing at all. @ClaudeDevs needs to answer this. We have active subscription
English
1
0
3
302
Sorower H.
Sorower H.@sorower01·
It's 18 July, and Fable is already removed from Claude Max subscriptions. Which supposed to be available till 19. @AnthropicAI ??
Sorower H. tweet media
English
5
0
6
1.5K
Mehul Mohan
Mehul Mohan@mehulmpt·
if you built a tool that is borderline unethical, how do you market it asking for a friend
English
73
1
161
15.1K
Vijay
Vijay@imvijaymode·
@iuditg 19th Feb? Did you mean july?
English
1
0
0
27
Udit Goenka
Udit Goenka@iuditg·
Anthropic will be stupid to remove Fable 5 on 19th Feb and only leave all the users with Opus 4.8. The pressure to release Opus 5.0 is even higher now.
Kimi.ai@Kimi_Moonshot

Meet Kimi K3

English
5
1
13
1.3K
Vijay
Vijay@imvijaymode·
@mehulmpt Can you try to make a detailed video on Kimi K3 end to end . If possible, cover how Kimi is able to achieve this level of intelligence with less compute and funding whereas it took frontier labs months ?
English
0
0
1
254
Vijay
Vijay@imvijaymode·
@waitin4agi_ Do Kimi K3 testing end to end and drop your review
English
0
0
0
84
Varun Mayya
Varun Mayya@waitin4agi_·
Blows my mind that most of society tried AI in 2023 and then stopped. They haven’t even tried Codex or Claude code yet. For people who stopped using AI in 2023/24 - you’re in for a surprise, pls try.
English
61
15
1K
55.3K
Gajesh
Gajesh@gajesh·
@sriramk these numbers do look very interesting -- testing the model right now with few autoresearch loops (def seeing opus / 5.6 terra result, maybe not as good as 5.6 sol yet) source of the bench: mp.weixin.qq.com/s/V4xhEIy8xDXS…
Gajesh tweet media
English
1
0
4
1.7K
Sriram Krishnan
Sriram Krishnan@sriramk·
kimi k3 is a big moment with multiple implications for the entire industry.
English
44
134
2K
148.2K
Paul Couvert
Paul Couvert@itsPaulAi·
@ECLresearch Correct. Looks like the Chinese labs are really good at inference optimization. DeepSeek, Kimi, etc... or western labs are overpricing APIs 👀
English
1
0
0
320
Paul Couvert
Paul Couvert@itsPaulAi·
Highly unexpected Kimi K3 is an open model already competitive with Fable 5 and 5.6 Sol... And even better on some tasks!!! - 2.8T parameters - 1M context window - Frontier on coding/agentic Last but not least: K3 has the same pricing as Sonnet 5 🤯 So you have ~ Fable 5/GPT-5.6 Sol intelligence but MUCH cheaper. First in frontend design arena as well! Open source keeps winning
Paul Couvert tweet media
Kimi.ai@Kimi_Moonshot

Meet Kimi K3

English
18
16
85
15.1K
Vijay
Vijay@imvijaymode·
@buildwithsid Agreed . Fable 5 is nowhere near to Kimi K3's output on frontend designs . Didn't test how its on backend engineering .
English
1
0
2
120
Vijay
Vijay@imvijaymode·
@adxtyahq What was the prompt? Is it real? 🤯
English
3
0
6
4.3K
aditya
aditya@adxtyahq·
Kimi K3 feels like China just open-sourced a Fable-5 class model Never seen an AI generate UI with this level of polish from a single prompt.
Kimi.ai@Kimi_Moonshot

Introducing Kimi K3: Open Frontier Intelligence 🔹 2.8 Trillion Parameters, 1 Million Context, Native Multimodal 🔹 Kimi Delta Attention enables up to 6.3x faster decoding in million-token contexts 🔹 Attention Residuals deliver ~25% higher training efficiency at <2% additional cost 🔹 Built for long-horizon agentic coding and self-evolving workflows Kimi K3 is now live on on Kimi.com, Kimi Work, Kimi Code, and the Kimi API. Open Weights by July 27, 2026. 🔗 API: platform.kimi.ai 🔗 Tech blog: kimi.com/blog/kimi-k3

English
59
102
2K
215.7K
ClaudeDevs
ClaudeDevs@ClaudeDevs·
Claude Code's /code-review now has effort levels, with the review rewritten at every one. Low effort beats other code review tools on findings at a fraction of the token cost. High effort delivers significantly higher recall when you want to go deeper. You pick the tradeoff.
ClaudeDevs tweet media
English
252
248
4.4K
541.5K
Vijay
Vijay@imvijaymode·
@flavioAd Didn't update the app 😛
English
0
0
1
280
Flavio Adamo
Flavio Adamo@flavioAd·
GPT 5.6-Sol Ultra no longer supports Full Access in Codex
Flavio Adamo tweet media
English
26
7
625
38.8K
Tushar Ojha
Tushar Ojha@techtusharojha·
GPT-5.6 Sol has been going on for 61 minutes. ⚠️ That too for an experiment (not a feature). I just gave a big self-written markdown file for an experiment on my product that I think might yield some interesting findings. Waiting excitedly to see how it goes (will post in this 🧵)
English
16
3
45
25.6K

Keşfet

@dee_bosa @iuditg @thsottiaux @maxedapps @AnthropicAI @balajis @claudeai @GladstoneInst