Mok

İndie useful insights. knowledge together!

[Mok @pratix01]

@elonmusk Will Grok V9-Medium be much stronger at real coding-agent workflows, like multi-file repo edits and debugging inside Cursor?

[Mok @pratix01]

Tavily got acquired. The market needs something sharper.
Soon. https://t.co/LT4RLBOu7k

#BuildInPublic #AIAgents #DevTools

[Mok @pratix01]

Yes

[Hridoy Rehman @hridoyreh]

Reddit is now losing SEO traffic:

[Mok @pratix01]

@_heyrico Are you sure these are the same font and it's a mistake?

[Mok @pratix01]

@HowToAI_ You're right, but this also burns LLM tokens, so it's not cost-effective, meaning there's a trade-off.

[Mok @pratix01]

Dostum sen konuyla alakalı konuştuğundan eminmisin yoksa çorba bol malzemeli olsun diye alakasız kavramlarmı yerleştiriyorsun “ontolojik köken” :)

[Fy @fatihynn]

İşte 'sığlık' tam olarak budur. 'Agentic OCR' dediğin şey metni okuyan bir paspastır; üç tane API'yi birbirine bağlayıp adına "Agentic" demek. Hal böyle olunca, "state-space" veya "ontolojik köken" gibi derin mühendislik kavramlar, Out of Memory (OOM) hatası verdiriyor. Sizin Agentic OCR dediğiniz şey metni okur ve unutur, accountability?? kanıt?

[Mok @pratix01]

@siddharthvij_ Very good bro

[Mok @pratix01]

Reallu beatiful :)

[Babayanski @Babayanski_12]

We always praise Tesla’s car design, but their app deserves just as much credit.ed their app actually is

[Mok @pratix01]

Beatiful design

[Babayanski @Babayanski_12]

80% of SaaS dashboards aren’t designed for clarity.

They just exist. That’s the problem.

This is what clarity looks like.

[Mok @pratix01]

That's crazy :)

[Kimi.ai @Kimi_Moonshot]

Meet Kimi K2.6 Agent Swarm 👋

Highlights：
🔹 Swarms, elevated - 300 parallel sub-agents × 4,000 steps per run (up from 100 / 1,500 in K2.5).

🔹 Outputs are real files, not chat - one run delivers 100+ files, 100,000-word literature reviews, or 20,000-row datasets.

🔹Heterogeneous skills - search, analysis, coding, long-form writing, and visual generation all running in parallel

🔗Try it at: https://t.co/2Tu8McUaUa

[Mok @pratix01]

Haha anlaşılmak isteyen ama olabildiğince anlaşılmaz yazan biri gibisin :) düz ocr yerine bir agentic (ve doğrulanabilir) bir zeka katmanı olabilirdi demek için çorba yapmışsın cümleleri bende acaba yabancı dilde yazdıda grok cümleleri anlaşılmazmı çevirdi diye kontrol ediyorum 😀

[Mok @pratix01]

@laibacreatives This is really nice :)

[Mok @pratix01]

Evet cidden böyle konuşunca güzel onu productionds çekince asıl mevzu başlıyor hele scale konusunu konuşmuyorum file bundan dolayı 2 hafta tam mesaji çalıştığım projeyi (kolay scale edenilmek için rust kullanmama rağmen) iptal etmek zorunda kaldım

[Mok @pratix01]

I think you should add it to version 2.6, then I’ll buy it :)

[BytePlus @BytePlusGlobal]

GLM-5.1 is now on BytePlus ModelArk Coding Plan. Starting at just $10/month, ModelArk Coding Plan offers a highly cost-efficient way to access GLM-5.1 alongside other advanced coding models.

GLM-5.1 is https://t.co/Jz7zvIeBkM's latest flagship model, MIT-licensed, open-weight, and built for long-horizon agentic coding. GLM-5.1 ranks among the world's top-tier models across leading coding benchmarks, including SWE-Bench Pro.

What you get with ModelArk Coding Plan:

→ Multiple advanced coding models in one subscription: GLM-5.1, Kimi-K2.5, Dola-Seed-2.0-pro, DeepSeek-V3.2, and more. Switch freely or let Auto mode match the best model to the task.

→ Works with the tools you already use: Claude Code, Cursor, Cline, Codex CLI, Kilo Code, Roo Code, OpenCode, and OpenClaw

→ No throttling. Backed by ByteDance's infrastructure.

→ Activated on purchase. Ready to use immediately.

Also new this month: Dreamina Seedance 2.0 is now available on BytePlus, the official API platform for Seedance models. Learn more: https://t.co/n5obvmuzgQ

Refer friends and earn 10% vouchers on every order with no cap. Your friends get 10% off their first subscription too.

Get started for $10/month → https://t.co/tN1toE3FjP

#BytePlus #ModelArk #GLM #AIEngineering #DevTools #AIAgent

[Mok @pratix01]

It looks nice :)

[Cristian Córdova 🐧 @barckcode]

Ayer en la comunidad me recomendaron @orca_build y es UNA PASADA. Está 100% pensado para agentes.

Es como si una terminal y un IDE se hubiesen casado y hubiesen tenido el hijo enfocado 100% en correr agentes. Está muy muy top.

10/10 la verdad.

[Mok @pratix01]

Fastest open source Ocr, I think you should give it a try :)

https://t.co/zGsbh79pGN

[Mok @pratix01]

I definitely don’t agree—it’s still not reliable enough for a full-scale project. You’ve probably just stumbled upon an area where it works well; I used K2.6 last night for a project; yes, it’s smarter, but when developing ReactFlow, it got stuck in a loop and couldn’t resolve the state inconsistency; Claude Code (with Opus 4.7), on the other hand, resolved it in one shot

[Mok @pratix01]

I definitely don’t agree—it’s still not reliable enough for a full-scale project. You’ve probably just stumbled upon an area where it works well; I used K2.6 last night for a project; yes, it’s smarter, but when developing ReactFlow, it got stuck in a loop and couldn’t resolve the state inconsistency; Claude Code (with Opus 4.7), on the other hand, resolved it in one go.

[Bindu Reddy @bindureddy]

Kimi K2.6 is Opus 4.7 level in agentic coding 🤯

We have got the receipts - will publish soon

[Mok @pratix01]

I definitely don’t agree—it’s still not reliable enough for a full-scale project. You’ve probably just stumbled upon an area where it works well; I used K2.6 last night for a project; yes, it’s smarter, but when developing ReactFlow, it got stuck in a loop and couldn’t resolve the state inconsistency; Claude Code (with Opus 4.7), on the other hand, resolved it in one go.

[Felipe Demartini @namcios]

Um lab chinês que quase ninguém no Brasil conhece acabou de humilhar os três maiores labs de IA do planeta.

Modelo open-source.
Pesos no HuggingFace. Gratuito.

E bate Claude Opus 4.6, GPT-5.4 e Gemini 3.1 Pro em 6 benchmarks.

Não é exagero.

A Moonshot lançou o Kimi K2.6 hoje:

→ SWE-Bench Pro: 58,6 (Claude: 57,7)
→ Toolathlon: 50,0 (Claude: 47,2)
→ SWE-Bench Multilingual: 76,7
→ BrowseComp: 83,2
→ HLE com tools: 54,0
→ MathVision com Python: 93,2

Agora a parte que deveria tirar o sono de toda big tech americana: o preço.

Kimi K2.6 via API: $0,60/milhão de tokens de input. $2,50 de output.
Claude Sonnet 4.6: $3,00 e $15,00.

5x mais barato no input. 6x no output.

E como os pesos são abertos, qualquer empresa com GPUs roda sem pagar nada para a Moonshot.

Mas o número mais assustador não é benchmark nem preço. É velocidade de execução.

O modelo rodou 4.000+ tool calls em uma sessão única. 12 horas de execução contínua. 300 sub-agentes em paralelo. Pegou um modelo local, reescreveu a inferência inteira em Zig, e foi de 15 tokens/segundo para 193. Sozinho.

Um engenheiro de software autônomo que trabalha 12 horas sem parar e não cobra salário. Open-source.

A OpenAI cobra $200/mês pelo Pro.
A Anthropic levantou $60 bilhões em valuation.
O Google queima $75 bilhões por ano em infraestrutura.

E um lab de Pequim, com uma fração desse capital, está entregando de graça o que essas empresas dizem aos investidores que custa dezenas de bilhões para construir.

A cadência é o que mata.

K2 em julho de 2025.
K2.5 em janeiro de 2026.
K2.6 agora.

A cada 8 semanas a Moonshot solta um modelo que come mais um pedaço do moat dos labs fechados. Dessa vez, em benchmarks agênticos, o moat evaporou.

Em janeiro o DeepSeek evaporou $600 bilhões da Nvidia em um único dia e forçou a OpenAI a tornar o ChatGPT gratuito na mesma semana.

Agora a Moonshot fez de novo.

Essa é a segunda vez em quatro meses. Vai ter uma terceira.

[Mok @pratix01]

I definitely don’t agree—it’s still not reliable enough for a full-scale project. You’ve probably just stumbled upon an area where it works well; I used K2.6 last night for a project; yes, it’s smarter, but when developing ReactFlow, it got stuck in a loop and couldn’t resolve the state inconsistency; Claude Code (with Opus 4.7), on the other hand, resolved it in one go.

[Erick @ErickSky]

El Kimi K2.6 es básicamente un Opus 4.7 más barato.

Es hora de activar ese plan.