Alan Puntim

[santi @santtiagom_]

les dejo uno de los mejores papers que leí sobre Harness Engineering.

explica muy bien:

1) qué es un harness
2) cómo están construidos los coding agents actuales
3) qué patrones se repiten entre Claude Code, Codex, Gemini CLI, etc.
4) hacia dónde está evolucionando la categoría
5) qué recomiendan si querés construir uno

todo parte de una definición simple:

Agent = Model + Harness

como complemento, también pueden leer el artículo que escribí sobre el tema.

paper:
https://t.co/xTvb7MvzQG

[Alan Puntim @arpuntim]

@0xCVYH @AkitaOnRails https://t.co/bFkSrEc9NA

[Alan Puntim @arpuntim]

Modelos open source estão apenas 4 meses atras dos modelos de fronteira americanos. https://t.co/naoVPnGOy4

[Alan Puntim @arpuntim]

@0xCVYH @culturabuilder Caminho sem volta, Major! 🚀🚀🚀

[Zest Protocol @ZestProtocol]

Bitcoin Collateral Vaults demo is now on mainnet.

Deposit native BTC into a self-custodial vault on Bitcoin L1. Borrow Ethereum USDC against it. The capital layer for Bitcoin.

[Mia @MiaAI_lab]

Come check out my website.

It includes a list of my latest & greatest recipes.

The "Install Model" button gives you a ready-to-use prompt for Claude, Codex, or others. It adapts the recipe to your setup, starts the server and tests it.

https://t.co/L9iTvPMF2k

[Rahul @sairahul1]

Codex tip: A cost-efficient DeepSeek V4.1 Flash agents, orchestrated by Astra.

Astra handles planning and review. Flash handles almost everything else at a fraction of the cost.

Hand this to Codex to set it up 👇

[Miguel Ángel Durán @midudev]

IA local según la memoria VRAM de tu máquina:

4 GB
└─ GPU integrada → Gemma 4 E2B Q4_K_M

8 GB
└─ RTX 5060 / 4060 → Gemma 4 12B QAT Q4_0

16 GB
└─ RTX 5060 Ti / 4080 → Gemma 4 12B Q8_0
└─ (al límite) Gemma 4 26B A4B Q4_K_M

24 GB
└─ RTX 3090 / 5080 Super → Qwen3.8-27B Q4_K_M

32 GB
└─ RTX 5090 → Qwen3.8-27B NVFP4 (velocidad) o Q6_K (calidad)
└─ M6 → Qwen3.8-27B Q4_K_M

48 GB
└─ M5 Pro → Qwen3.8-27B Q8_0

64 GB
└─ 2× RTX 5090 → Qwen3.8-27B BF16
└─ M5 Pro → Qwen3.8-27B BF16

96 GB
└─ RTX PRO 6000 Blackwell → Qwen3.8-Flash-Next UD-Q2_K_XL
└─ M5 Ultra → Qwen3.8-Flash-Next UD-Q3_K_XL

128 GB
└─ DGX Spark → Qwen3.8-Flash-Next UD-Q4_K_XL
└─ M5 Max → Qwen3.8-Flash-Next UD-Q3_K_XL

256 GB
└─ M5 Ultra → DeepSeek V4 Flash 0731 UD-Q8_K_XL
└─ 2× DGX Spark → DeepSeek V4 Flash 0731 UD-Q8_K_XL

[Alan Puntim @arpuntim]

@0xCVYH Estou criando um fork do DSH para minha empresa.

[Kirill @kirillk_web3]

A Chinese developer just explained the shift from Loop Engineering to Graph Engineering better than anyone.

most people are still building agents the way that's about to be obsolete.

> why single-agent loops break and go "goal blind"
> the 4 parts of a graph: nodes, edges, state, policy
> 3 topologies that run everything: diamond, supervisor, pipeline
> Anthropic's 5 official workflow patterns

the punchline: it's not how many agents you run. it's the determinism you build with verifiers, code fallbacks, and reality anchors.

I broke the same architecture down with Kimi K3. Full A-Z guide below.

[Alan Puntim @arpuntim]

@0xCVYH To rodando no mac mini de 24gb Q4 ta rodando muito bem. Detalhe que esta rodando com 11 token/s lerdo, uso apenas para tarefas autonomas, que a skill já foi validada com modelos de fronteira. Porem resolve. Outro detalhe, pensa muito. Estudando como achar a melhor configuracao.

[Alan Puntim @arpuntim]

@ojedabtc Agora, -1000

[Roan @RohOnChain]

My friend applied to 200 tech jobs in two years. No MIT. No Stanford.

Last month Anthropic offered him $750,000.

I asked him how he broke in from zero.

He sent me the exact video that got him in. Anthropic's 2 hour course on "How to become an AI engineer in 2026"

Their core team shows exactly how to architect & build AI agents from scratch.

I watched it last night.

Halfway through, I realized I could break into an AI lab in weeks, not years.

Bookmark this and read the article below.

• 00:00 - AI agents with graph engineering
• 06:41 - AI agent architecture
• 14:31 - building AI agent loops live
• 1:15:24 - AI agentic RAG
• 2:20:42 - Anthropic interview process

[Crypto Tice @CryptoTice_]

LONG TERM HOLDERS ARE FLOODING EXCHANGES WITH BITCOIN.

5.1% of all exchange inflows coming from the most convicted cohort in this market.

Near record levels.
Only 2020 saw higher at 5.5%.

These are not weak hands breaking.
These are the strongest hands in the market making a deliberate move.

The last time LTH exchange inflows hit these levels...
Bitcoin was about to begin a 700% rally.

The smartest money is active right now.
The only question is which direction they are moving.

[Gabriel Santos @DehumanoaDeus]

Em vez de passar uma hora vendo Netflix, assista a esta aula de 2 horas de Stanford.
Ela vai te ensinar mais sobre como LLMs como ChatGPT e Claude são criados do que a maioria das pessoas que trabalham nas maiores empresas de IA aprende durante toda a carreira.

[Alan Puntim @arpuntim]

@0xCVYH Major envia os @ dos caras gringos que vc acompanha.

[Alan Puntim @arpuntim]

@0xCVYH Bora comprimir esse modelo pra ontem!

[Circle @circle]

Circle Wrapped Bitcoin is coming.

Backed 1:1 by BTC and readily verifiable onchain, cirBTC is being built to work seamlessly with Circle infrastructure and the broader DeFi ecosystem.

Learn more: https://t.co/wWzVBZdIz1

[FELPS @FelpsCrypto]

🚨 ATENÇÃO: O Claude consegue te ensinar qualquer idioma como um professor particular de R$500/hora da Berlitz. De graça.

Aqui estão 12 prompts que te deixam conversacional em qualquer idioma em 30 dias:

(Salva antes de sumir)