Dipesh
1.6K posts
Dipesh
@dipeshdotdev
Backend | Distributed Systems | Infra Indie Builder shipping in public https://t.co/OamRdBABbK | https://t.co/rB4ShrV5gi | https://t.co/ivRMIyvfvv
India Katılım Kasım 2020
124 Takip Edilen202 Takipçiler
Someone posted this and I'm all the way ☠️☠️
Fynn@fynnso
was messing with the OpenAI base URL in Cursor and caught this accounts/anysphere/models/kimi-k2p5-rl-0317-s515-fast so composer 2 is just Kimi K2.5 with RL at least rename the model ID
English
🚨NEWS: Cursor’s $50B “in-house model” is literally Kimi K2.5 with RL on top. Got caught in 24 hours
>be Moonshot AI
>spend hundreds of millions training Kimi K2.5
>1 trillion parameters, 15 trillion tokens, agent swarm architecture
>beat GPT-5.2 and Opus 4.5 on real benchmarks
>open-source it because you believe in the ecosystem
>one condition: display “Kimi K2.5” if you make over $20M/month from it
>Cursor takes the model
>runs RL on coding tasks
>ships it March 19 as “Composer 2”
>blog post: “continued pretraining + scaled reinforcement learning”
>zero mention of Kimi K2.5
>“our in-house models generate more code than almost any other LLMs in the world”
>publishes benchmark chart
>Composer 2 against Opus 4.6 and GPT-5.4
>uses the chart to justify raising at $50 billion!
>less than 24 hours later
>kimi dev intercepts the API response
>model ID: kimi-k2p5-rl-0317-s515-fast
>they didn’t even rename it
>Moonshot head of pretraining runs tokenizer test
>confirms: identical to Kimi’s tokenizer
>publicly tags Cursor’s co-founder: “why aren’t you respecting our license?”
>two more Moonshot employees post confirmations
>all three posts deleted within hours
>legal is now involved
>but it gets worse
>Cursor had Kimi K2.5 listed as a FREE model in their UI just weeks ago
>users were openly using it
>Feb 9: “K2.5 was in my model list. I updated and it vanished”
>it vanished because Cursor pulled it from the picker, and relaunched it as their own model
>Moonshot valuation: $4.3B
>Cursor valuation: $50B
Absolute state of Cursor.
Elon Musk@elonmusk
@fynnso Yeah, it’s Kimi 2.5
English
Breaking down the problem further:
Right now everything is fragmented:
→ server metrics → one tool
→ logs → another
→ containers → CLI
→ compose grouping → manual
And the worst part:
There’s no clear unit to think in.
But in reality, we think like this:
→ Server
→ Services (compose apps, independent processes, solo containers)
→ Containers
That became the core model.
Not metrics-first.
Not logs-first.
But:
→ server → service → container
A model which is more natural for small teams and solo devs.
Once this clicked, the product direction became a bit clear now.
Open to your suggestions.
x.com/dipeshdotdev/s…
English
@oliviscusAI this is just looking awesome man !!! looking at things from audiance pov is indeed a good way to learn about systems.
English
someone built a web-based System Design Simulator.
you drag and drop components (api gateways, dbs, caches) and it actually simulates real-time traffic.
you can watch latency, bottlenecks, and failures happen live...
English
I run multiple VPS with Docker-heavy workloads.
Setup looks like:
→ Postgres, Redis, Nginx
→ multiple compose apps (APIs, workers, frontends)
→ local LLM services (Whisper, Ollama)
Monitoring this is painful.
Not because tools don’t exist.
But because they’re built for a different scale.
→ Grafana + Loki = powerful but heavy
→ setup takes time
→ overkill for small teams
So I asked:
What do I actually need daily?
→ what’s running?
→ is it healthy?
→ where are the logs?
That’s the real problem. So thinking of a building a simple solution around it.
English
came across a senior dev’s portfolio
using linux since 1999
27 years experience
C/C++ (27yr), JS (24yr), Python (20yr), Java (22yr), Go, Rust…
bro has more experience in one language than most of us have been alive
whole different league
portfolio -> kamilnowicki.com
English
@thesayannayak I don't know sometimes I am not allowed to reply or post at all, because it thinks its a bot trying to perform the operation. like fr men.
English
@VANIXQBIT_LLP @gregisenberg Yeah I touch grass, spend quality time with friends and family often now.
English
reminder that you’re not behind, it’s just moving too fast
English
@itsmewhocodes @gregisenberg Codex is available in go as well and the limits are good for small projects. Give it a try.
English
@dipeshdotdev @gregisenberg I want to be in top 0.3 but paise nh h bhaiya aap claude pro and codex k subscription dedo , mera bhi kaam aasan hojaayega thoda bhaiya
Eesti
@ravikiran_dev7 just ask the user to use a vpn or they can travel to england and then it will be loaded in 100 ms.
when can I join?
English
Dipesh retweetledi
I was away from Twitter for a week.
Nobody in real life talked about AI. Not once.
No one mentioned Claude, GPT-5, or whatever startup raised millions.
People knew Meta AI because it's in WhatsApp. They knew ChatGPT because it went viral. That's it.
Meanwhile here, every tweet is "AI will replace you" or "I built this in 5 minutes".
Creates insane pressure like you're falling behind every day.
But look at this graph.
84% of people have NEVER used AI.
16% use free chatbots.
0.3% pay for AI tools.
If you're using premium AI tools, you're in the top 0.3% globally.
The "AI race" only exists on Twitter and Silicon Valley.
Most people don't even know these tools exist.
So stop feeling like you can't keep up.
Twitter makes you feel behind. Reality shows you're way ahead.
English
Also I am not able to connect codex with github as well.
Dipesh@dipeshdotdev
Trying to upgrade to @OpenAI pro plan since it will have higher usage limit for codex, but the process is so frustrating. I guess they might be losing their customer just before the payment and they might never know about it. I am getting this same error everytime. I am not even getting an option to choose the payment method. If it's using the default method then why is it not working with OpenAI when it works with other services? The error is so generic. I have checked my card twice for the limit and I have also enabled international payments, still it's not working.
English
Trying to upgrade to @OpenAI pro plan since it will have higher usage limit for codex, but the process is so frustrating. I guess they might be losing their customer just before the payment and they might never know about it.
I am getting this same error everytime. I am not even getting an option to choose the payment method. If it's using the default method then why is it not working with OpenAI when it works with other services?
The error is so generic. I have checked my card twice for the limit and I have also enabled international payments, still it's not working.
English
we didn’t think it was possible
after multiple years of R&D, the technology is finally here
Notion@NotionHQ
Heading 4 is finally here 😤 The years of “just bold the text and pretend” are over. Rolling out now.
English
I have not decided yet but the arch decision is done, I will build a simple go agent which will run on server and will send data to web app. I have kept it stupid simple, this solution might not be for big enterprises with complex systems but this will really be good for solo devs and small teams. Stay tuned I should be able to create a deployed mvp in a week.
English
@dipeshdotdev The complexity problem is real. Most teams run Prometheus + Grafana + Loki + alertmanager just for what you described — before on-call or a status page. "Deploy a service, instantly see everything" UX is genuinely missing. What are you using for metrics collection?
English
Today I’m starting work on a new project → Bifrost
The goal is simple:
Modern observability tools are too complex
→ too many components
→ too much setup
→ too hard to maintain
And for small setups (like VPS + Docker apps), it feels like overkill.
Right now, if I want to:
→ monitor services
→ check logs
→ see system stats
I have to use multiple tools.
That’s the problem I want to solve.
Bifrost will be:
→ lightweight
→ easy to set up
→ focused on real needs
→ metrics + logs in one place
If I deploy a new service tomorrow,
I should instantly see:
→ its resource usage
→ its logs
→ everything in one clean UI
No complexity. No friction.
PRD is done.
Next step → start building.
English