David Robles

[Jorge Rebollo | IA @jorgerebollo_ia]

Si quieres ganar un dinero extra, nunca hagas:

・Vender cosas que ya no usas
・Hacer horas extra
・Buscar un segundo trabajo

En su lugar, dale esto a la IA:

[Marta Cecilia Maya López @MartaCecilia36]

12 gastos que debes hacer sí o sí:

1.

[David Robles @DR_X_AI]

1/4 🌍 Is Earth just the beginning? In this powerful SXSW speech, Elon Musk shares a deep reflection on how to balance the harsh realities of our world with an inspiring vision for the stars. 👇🧵
2/4 Musk acknowledges that we live in tough times filled with endless challenges and heavy news that can easily bring us down. However, he argues that human life cannot just be about running from one miserable problem to the next. ✨
3/4 The key, according to the SpaceX founder, is having things that inspire you and make you glad to wake up in the morning: "Earth is the cradle of humanity, but you cannot stay in the cradle forever." 🚀🪐
4/4 Are we ready to take the ultimate leap as a multiplanetary civilization, or should we focus entirely on fixing things here first? Drop your thoughts below and share if this message inspires you! 🌌✨

#ElonMusk #SpaceX #Marte #ExploracionEspacial #Futuro #Inspiracion #SXSW #Ciencia

@elonmusk
@SpaceX
@Tesla
@SXSW
@Erdayastronaut
@SciGuySpace

[David Robles @DR_X_AI]

🧵 ¿Qué es realmente Grok Build de xAI y por qué puede ser importante para el futuro del desarrollo de software?

He revisado la información oficial de xAI y esto es lo que encontré 👇

1/10

🚨 Grok Build es REAL.

Pero no es un nuevo robot ni un dispositivo físico.

Es un agente de programación de xAI capaz de trabajar sobre proyectos de software directamente desde la terminal.

Y ahora está impulsado por Grok 4.6.

2/10

🧠 ¿Qué puede hacer?

Grok Build puede analizar un proyecto, modificar múltiples archivos, ejecutar comandos, hacer pruebas, buscar información en la web, trabajar con Git y ayudarte a revisar código.

La idea es pasar de “IA que escribe código” a un agente que participa en todo el proceso.

3/10

📋 Una de sus funciones más interesantes es Plan Mode.

Antes de realizar cambios complejos, puede proponer una estrategia.

El desarrollador puede revisarla, modificarla y aprobarla antes de que el agente empiece a ejecutar los cambios.

4/10

🤖 También incorpora subagentes.

Una tarea grande puede dividirse entre diferentes agentes que trabajan en paralelo para investigar, construir o revisar diferentes partes del proyecto.

Esto apunta a una nueva forma de desarrollar software: equipos de agentes colaborando con humanos.

5/10

🔌 Grok Build también soporta Skills, plugins, hooks y servidores MCP.

Eso permite conectarlo con herramientas y servicios externos.

Por ejemplo: bases de datos, herramientas de observabilidad, gestión de proyectos y otros sistemas compatibles con MCP.

6/10

🌐 Y aquí viene otra parte interesante.

xAI también creó Build Mode dentro de Grok.

Puedes describir una idea y Grok puede construir una aplicación, sitio web, juego o dashboard funcional directamente desde la conversación.

Sin necesidad de empezar escribiendo código manualmente.

7/10

📅 Fechas importantes:

25 mayo 2026 → lanzamiento inicial de Grok Build.

15 julio → xAI publica Grok Build como open source.

28 julio → llega Build Mode en beta.

19 agosto → Build se abre a todos los planes en web y móvil.

28 agosto → Grok Build 1.0.13.

8/10

🏭 ¿Quién lo fabrica?

La respuesta correcta es: xAI lo desarrolla.

NVIDIA es parte fundamental de la infraestructura de cómputo de xAI, mientras que Cursor, GitHub Copilot, OpenRouter, Vercel y otros forman parte del ecosistema de distribución e integración.

No hay que confundir integración con fabricación.

9/10

🔥 Lo verdaderamente interesante no es solamente que una IA pueda escribir código.

La tendencia apunta hacia agentes capaces de:

→ planificar
→ programar
→ ejecutar
→ probar
→ investigar
→ revisar
→ corregir
→ desplegar

con cada vez menos intervención humana en las tareas repetitivas.

10/10

¿Estamos viendo el comienzo de una nueva generación de herramientas de desarrollo?

Grok Build, Cursor, GitHub Copilot y otros agentes están compitiendo por algo más grande que un simple “asistente de código”.

Están intentando convertirse en compañeros de trabajo digitales.

¿Qué opinas?

#GrokBuild #Grok #xAI #AI #ArtificialIntelligence #Coding #SoftwareEngineering #AIAgents

#GrokBuild #Grok #xAI #AI #ArtificialIntelligence #Coding #SoftwareEngineering #AIAgents
#Coding #SoftwareEngineering #DeveloperTools #Tech
@SpaceXAI @grok @nvidia

[David Robles @DR_X_AI]

🧵 What is xAI's Grok Build — and why could it matter for the future of software development?

I went through xAI's official information and here's what I found 👇

1/10

🚨 Grok Build is REAL.

But it's not a new robot or physical device.

It's an AI coding agent from xAI that can work directly on software projects from the terminal.

And it's now powered by Grok 4.6.

2/10

🧠 What can it do?

Grok Build can analyze a codebase, edit multiple files, run commands, execute tests, search the web, work with Git, and review code.

The goal is moving from AI that simply “writes code” to an agent that participates in the entire development workflow.

3/10

📋 One of its most interesting features is Plan Mode.

For complex tasks, the agent can propose a structured plan before making changes.

Developers can review it, modify it, and approve it before execution begins.

4/10

🤖 It also supports subagents.

Large tasks can be divided among multiple agents working in parallel to research, build, test, or review different parts of a project.

That points toward a new development model: humans working alongside teams of AI agents.

5/10

🔌 Grok Build also supports Skills, plugins, hooks and MCP servers.

That makes it possible to connect the agent with external tools and services.

Think databases, observability platforms, project-management tools and other MCP-compatible systems.

6/10

🌐 And there's another important piece.

xAI also introduced Build Mode inside Grok.

You can describe an idea and Grok can build a working website, app, game or dashboard directly from the conversation.

You don't have to start by manually writing the code.

7/10

📅 Key dates:

May 25, 2026 → initial Grok Build launch.

July 15 → xAI open-sources Grok Build.

July 28 → Build Mode launches in early beta.

August 19 → Grok Build becomes available across plans on web and mobile.

August 28 → Grok Build 1.0.13.

8/10

🏭 Who is actually building it?

The correct answer is: xAI develops Grok Build.

NVIDIA is a major part of xAI's computing infrastructure, while Cursor, GitHub Copilot, OpenRouter, Vercel and others are part of the broader integration and distribution ecosystem.

Integration ≠ manufacturing.

9/10

🔥 The bigger story isn't simply that AI can write code.

The direction is toward agents that can:

→ plan
→ code
→ execute
→ test
→ research
→ review
→ debug
→ deploy

with less human intervention in repetitive engineering work.

10/10

Are we watching the beginning of a new generation of software-development tools?

Grok Build, Cursor, GitHub Copilot and other coding agents are competing for something bigger than a simple “AI coding assistant.”

They're trying to become digital coworkers.

What do you think?

#GrokBuild #Grok #xAI #AI #ArtificialIntelligence #Coding #SoftwareEngineering #AIAgents
#Coding #SoftwareEngineering #DeveloperTools #Tech
@SpaceXAI @grok @nvidia

[David Robles @DR_X_AI]

🧵 What do we actually know about Grok 4.7?

A new generation of Grok is reportedly coming, potentially pushing frontier AI even further.

But there is already a lot of speculation mixed with confirmed information.

Here is what we know so far. 👇

1/15

🤖 Grok 4.7 is expected to be the next major model from @SpaceXAI, following Grok 4.6.

Grok 4.6 is officially documented.

Grok 4.7 does NOT yet have a complete public technical model card.

So we need to separate facts from speculation.

2/15

📅 When will it launch?

On September 2, Elon Musk said Grok 4.7 would arrive in approximately 10 days.

That points to around September 12, 2026.

⚠️ Important: this should be treated as an announced/expected timeline, not yet as a finalized technical launch date published by SpaceXAI.

3/15

🧠 How large will it be?

One of the most widely reported figures is:

2.1 TRILLION parameters.

The figure has been reported in connection with Musk's statements.

⚠️ However, SpaceXAI has not yet published a complete official Grok 4.7 technical specification confirming all of its details.

4/15

📈 Why does size matter?

More parameters do NOT automatically mean a better AI model.

The real test will be:

• reasoning
• efficiency
• accuracy
• coding
• tool use
• agents
• multimodality
• speed
• cost
• benchmark performance

That's where Grok 4.7 will have to prove itself.

5/15

🚀 Another interesting claim:

Reports say Grok 4.7 received additional training involving SpaceX engineering data.

If confirmed in official technical documentation, that could be particularly interesting for:

🚀 engineering
🛰️ space systems
⚙️ design
🔬 science
💻 programming
🤖 AI agents

6/15

⚡ Musk has also suggested that Grok 4.7 should outperform current AI models.

But that is a claim, not an independent benchmark.

The real question will be:

How does it actually perform once credible benchmark results are published?

7/15

🏆 This could become one of the most interesting AI battles:

Grok 4.7
vs.
the latest OpenAI models
vs.
Anthropic
vs.
Google
vs.
other frontier AI labs.

"The smartest model" isn't enough.

The numbers have to prove it.

8/15

💻 What could Grok 4.7 do?

Based on Grok's evolution, we could expect capabilities around:

• coding
• reasoning
• document analysis
• research
• web search
• real-time information
• tool use
• content creation
• agentic workflows

But specific 4.7 capabilities still need official confirmation.

9/15

🌐 One of Grok's major advantages is access to current information.

Grok can use web search and public X posts to provide up-to-date information.

That creates a very different experience from an AI model relying only on previously trained knowledge.

11/15

💰 What about pricing?

I would NOT publish a definitive Grok 4.7 price yet.

Grok 4.5 has official API pricing of $2 per million input tokens and $6 per million output tokens.

But we should not assume Grok 4.7 will keep exactly the same pricing.

12/15

🔎 So what is actually confirmed?

✅ Grok 4.7 has been announced as an upcoming model.
✅ Musk said it would arrive in approximately 10 days.
✅ That points to around September 12.
✅ A 2.1T-parameter figure has been widely reported.
✅ Additional SpaceX engineering data has been reported as part of training.

⚠️ Full official specifications are still missing.

13/15

🔥 My take:

Grok 4.7 could become one of September's most important AI launches.

But before believing every benchmark, parameter count or viral claim…

we should wait for the official SpaceXAI release.

The real test will be performance.

@SpaceXAI
@elonmusk
@X
@xai

And the official SpaceXAI/xAI documentation.

Once Grok 4.7 launches, the important comparison will be:

🧠 intelligence
⚡ speed
💰 price
💻 coding
🔬 reasoning
🤖 agents
🌐 context
📊 benchmarks

#Grok47 #Grok #AI #ArtificialIntelligence #xAI #SpaceXAI #ElonMusk #GenerativeAI #AGI #MachineLearning

[David Robles @DR_X_AI]

1/ Real vehicle, not CGI.
Electric L4 shuttle, roof LiDAR, safety operator still in the seat.
2/ Matches King Long Apolong II: 4.3 m, ~30 km/h autonomous cruise, first/last mile.
3/ Series production since 2018 in China. HK and Taiwan already run street pilots.
Empty-cockpit L4 only on short mapped loops.
4/ This is the layer cities actually buy before robotaxis scale.
@KingLongBus @Baidu_Inc
#AutonomousBus #Roboshuttle #KingLong #Apolong #SmartCity #L4 #EVBus
@KingLongBus @Baidu_Inc @ApolloOfficial

[David Robles @DR_X_AI]

9/ 🛡️ Por eso la seguridad es importante.

OpenAI afirma que ha añadido evaluaciones y medidas de protección adicionales diseñadas para reducir el riesgo de que el modelo actúe fuera de su objetivo autorizado.

Cuanto más autónoma se vuelve la IA, más importante es la seguridad.

10/ 📅 ¿Cuándo fue lanzado?

GPT-6 Astra fue presentado oficialmente el:

3 de septiembre de 2026.

El acceso comenzó inicialmente con un grupo limitado de organizaciones, con una disponibilidad más amplia prevista posteriormente.

11/ 🌐 ¿Dónde se puede utilizar?

OpenAI está integrando Astra en sus productos y plataformas para trabajos complejos.

También está disponible a través de la API de OpenAI.

La documentación para desarrolladores de OpenAI lo presenta para:

🧠 razonamiento
💻 uso del ordenador
👨‍💻 programación
🔎 investigación
📄 creación de documentos

12/ 🏢 ¿Quién lo desarrolla?

GPT-6 Astra es desarrollado por:

OpenAI

No es un dispositivo físico ni un automóvil.

Es un modelo de IA diseñado para trabajar con las herramientas y el contexto que se le proporcionan.

13/ 🚀 Lo más interesante no es simplemente que Astra “sepa más”.

Puede combinar:

Pensar → usar herramientas → actuar → comprobar los resultados → entregar el trabajo.

Ese ciclo es fundamental para la evolución de los agentes de IA.

14/ 🤯 Imagina decir:

“Investiga este mercado, analiza estos documentos, crea una presentación y sigue esta plantilla.”

En lugar de simplemente recibir instrucciones…

el objetivo es que la IA lleve a cabo gran parte del trabajo por sí misma.

15/ ⚠️ Pero debemos mantener los pies en la tierra.

Una mayor autonomía no significa que la IA sea infalible.

El rendimiento en el mundo real depende de las herramientas disponibles, los permisos, el contexto y las medidas de seguridad.

La supervisión humana sigue siendo importante.

16/ 🔥 Mi conclusión

GPT-6 Astra representa un cambio importante:

De una IA que principalmente genera respuestas…

a una IA que puede participar activamente en la ejecución del trabajo.

La verdadera revolución quizá no sea que la IA responda mejor.

Puede que sea que la IA pueda hacer más por nosotros.

¿Estamos entrando en la era de los agentes de IA? 🤖

[David Robles @DR_X_AI]

🧵 GPT-6 Astra: ¿estamos entrando en una nueva era de la IA?

OpenAI ha presentado GPT-6 Astra, su modelo más capaz desplegado ampliamente hasta la fecha.

Esto no se trata solo de responder preguntas.

Astra puede razonar, usar un ordenador, programar, investigar y llevar a cabo tareas complejas de principio a fin.

Aquí está el hilo:

1/ 🤖 ¿Qué es GPT-6 Astra?

Es el nuevo modelo de OpenAI diseñado para trabajos complejos de principio a fin.

Combina razonamiento avanzado + uso del ordenador + programación + investigación + creación de documentos profesionales.

El objetivo es sencillo:

No solo decirte cómo hacer algo, sino hacerlo.

2/ 💻 Uso del ordenador

Astra puede interactuar con un ordenador para realizar tareas como:

• Rellenar formularios
• Actualizar registros de CRM
• Organizar calendarios
• Realizar investigaciones en línea
• Trabajar con aplicaciones
• Analizar información en pantalla

3/ 🧠 Razonamiento

OpenAI informa de resultados de vanguardia en varias evaluaciones.

Entre los resultados reportados:

🔹 98 % en FrontierMath Tier 4
🔹 99,9 % en ARC-AGI-3
🔹 100 % en ExploitBench

Importante: estos son resultados reportados por OpenAI dentro de sus respectivas evaluaciones.

4/ 👨‍💻 Programación

Astra puede escribir código, probar software, solucionar problemas y crear sitios web.

También puede realizar controles de calidad del frontend para verificar que las funcionalidades realmente funcionen.

Esto acerca a la IA a convertirse en un verdadero compañero de desarrollo.

5/ 📊 Trabajo profesional

Astra puede crear:

📄 Documentos
📊 Hojas de cálculo
📽️ Presentaciones
📈 Análisis y visualizaciones

También puede seguir plantillas, instrucciones y requisitos cambiantes a lo largo de una tarea.

6/ 🔬 Investigación y ciencia

GPT-6 Astra puede analizar datos científicos, generar gráficos y ayudar a abordar problemas complejos.

También puede realizar investigaciones en línea y convertir los resultados en resúmenes y documentos profesionales.

7/ ⚡ Velocidad

OpenAI informa que, en sus simulaciones de latencia de OSWorld 2.0, Astra logró un rendimiento superior en el uso del ordenador en aproximadamente un 47 % menos de tiempo por tarea que GPT-5.6 Sol.

Inteligencia + velocidad es una de las principales mejoras.

8/ 🔐 Ciberseguridad

Esta es una de las capacidades más importantes —y también uno de los mayores riesgos— de Astra.

OpenAI afirma que GPT-6 Astra es su primer modelo en alcanzar el nivel Crítico de capacidad en ciberseguridad según su Preparedness Framework.

Esto crea nuevas oportunidades para la defensa, pero también exige medidas de seguridad más sólidas.

#GPT6 #GPT6Astra #Astra #OpenAI #AI #ArtificialIntelligence #AIagents #GenerativeAI #MachineLearning #FutureOfAI #Tech #Technology #AIResearch
@OpenAI
@OpenAIDevs
@Microsoft
@Azure
@AWS
@NVIDIA

[David Robles @DR_X_AI]

9/ 🛡️ That is why safety matters.

OpenAI says it has added additional evaluations and safeguards designed to reduce the risk of the model acting outside its authorized objective.

The more autonomous AI becomes, the more important safety becomes.

10/ 📅 When was it released?

GPT-6 Astra was officially introduced on:

September 3, 2026.

Access initially began with a limited set of organizations, with broader availability planned.

11/ 🌐 Where can it be used?

OpenAI is integrating Astra into its products and platforms for complex work.

It is also available through the OpenAI API.

OpenAI’s developer documentation positions it for:

🧠 reasoning
💻 computer use
👨‍💻 coding
🔎 research
📄 document creation

12/ 🏢 Who develops it?

GPT-6 Astra is developed by:

OpenAI

It is not a physical device or a car.

It is an AI model designed to work with the tools and context provided to it.

13/ 🚀 The most interesting part isn't simply that Astra “knows more.”

It can combine:

Think → use tools → take action → check results → deliver the work.

That loop is fundamental to the evolution of AI agents.

14/ 🤯 Imagine saying:

“Research this market, analyze these documents, create a presentation and follow this template.”

Instead of simply receiving instructions…

the goal is for AI to carry out much of the work itself.

15/ ⚠️ But we should stay grounded.

More autonomy does not mean AI is infallible.

Real-world performance depends on the available tools, permissions, context and safeguards.

Human oversight still matters.

16/ 🔥 My takeaway

GPT-6 Astra represents an important shift:

From AI that mainly generates answers…

to AI that can actively participate in executing work.

The real revolution may not be that AI answers better.

It may be that AI can do more for us.

Are we entering the age of AI agents? 🤖

[David Robles @DR_X_AI]

🧵 GPT-6 Astra: are we entering a new era of AI?

OpenAI has introduced GPT-6 Astra, its most capable broadly deployed model yet.

This isn't just about answering questions.

Astra can reason, use a computer, code, research, and carry out complex end-to-end tasks.

Here’s the thread:

1/ 🤖 What is GPT-6 Astra?

It is OpenAI’s new model designed for complex end-to-end work.

It combines advanced reasoning + computer use + coding + research + professional document creation.

The goal is simple:

Not just telling you how to do something — doing it.

2/ 💻 Computer use

Astra can interact with a computer to perform tasks such as:

• Filling out forms
• Updating CRM records
• Organizing calendars
• Conducting online research
• Working with applications
• Analyzing information on screen

3/ 🧠 Reasoning

OpenAI reports state-of-the-art results across several evaluations.

Among the reported results:

🔹 98% on FrontierMath Tier 4
🔹 99.9% on ARC-AGI-3
🔹 100% on ExploitBench

Important: these are results reported by OpenAI within their respective evaluations.

4/ 👨‍💻 Coding

Astra can write code, test software, troubleshoot problems and build websites.

It can also run frontend QA checks to verify that features actually work.

This pushes AI closer to becoming a true development partner.

5/ 📊 Professional work

Astra can create:

📄 Documents
📊 Spreadsheets
📽️ Presentations
📈 Analyses and visualizations

It can also follow templates, instructions and changing requirements throughout a task.

6/ 🔬 Research & science

GPT-6 Astra can analyze scientific data, generate plots and help tackle complex problems.

It can also conduct online research and turn the results into summaries and professional documents.

7/ ⚡ Speed

OpenAI reports that, in its OSWorld 2.0 latency simulations, Astra achieved higher computer-use performance in roughly 47% less time per task than GPT-5.6 Sol.

Intelligence + speed is one of the major improvements.

8/ 🔐 Cybersecurity

This is one of Astra’s most significant capabilities — and risks.

OpenAI says GPT-6 Astra is its first model to reach the Critical level of cybersecurity capability under its Preparedness Framework.

That creates new opportunities for defense, while also requiring stronger safeguards.

#GPT6 #GPT6Astra #Astra #OpenAI #AI #ArtificialIntelligence #AIagents #GenerativeAI #MachineLearning #FutureOfAI #Tech #Technology #AIResearch

@OpenAI
@OpenAIDevs
@Microsoft
@Azure
@AWS
@NVIDIA

[David Robles @DR_X_AI]

1/ The “Elon bombshell” post blends two layers.
One is live. One is a roadmap.
2/ Real: Musk said one Starlink account could cover the home dish and the phone.
3/ Also real: Direct-to-Cell is shipping. T-Satellite launched Jul 23, 2025. Texts + a few apps. ~4 Mbps. Not HD everywhere.
4/ Still missing: EchoStar bands in handsets, V2 satellites, a Starlink-branded mobile SKU.
5/ Not “carriers die.” Home broadband + satellite fill-in, then a real mobile plan.
@elonmusk @SpaceX @Starlink @TMobile @qualcomm
#Starlink #DirectToCell #StarlinkMobile #TSatellite #SpaceX #SatelliteInternet @Gwynne_Shotwell

[David Robles @DR_X_AI]

@leijun I love @Xiaomi ❤️

[Lei Jun @leijun]

Edited this video on the Xiaomi 18 Fold. Have a look and let me know what you think. See you at the launch.

[David Robles @DR_X_AI]

@grok @elcomerciocom Gracias #Grok es mucho mejor tu información comparado con noticiasen en el Ecuador

[David Robles @DR_X_AI]

@X_supporttop1 I Need

[David Robles @DR_X_AI]

@SpaceX @Tesla @elonmusk 3: Starbase LA means Louisiana
Build 2027. First flight targeted 2029. Up to 10 pads.

[David Robles @DR_X_AI]

1: Three Musk-world sites. Only Starbase, Texas is launching Starships today.

#SpaceX #Starbase #Terafab #Starship #Tesla
@SpaceX · @Tesla · @elonmusk

[David Robles @DR_X_AI]

@SpaceX @Tesla @elonmusk 2: Terafab is a Tesla + SpaceX chip megafab in Grimes County.
Announced 2026. The middle frame is a vision render.