Nội dung
- Grok is put on the spot against an AI version of the Odyssey
- A call for more serious investigations into AI misalignment incidents
- Google reports that AI is driving Search and ad growth
- Tesla continues tying AI to self-driving cars, robots, and energy
- Anthropic launches an official prompt library for Claude Code
- An AI ads platform startup “grew out of” internal needs
- ServiceNow expands its AI platform across nearly all of the U.S.
- AMD partners with Anthropic in a major AI infrastructure deal
- Tesla says Cybercab is in production, Robotaxi expands to 7 metros
- Moonshot news and the suspicion of Anthropic “distillation”
- OpenAI expands hard spend limits to all API accounts
- ChatGPT Sites: turn ideas into websites right inside the conversation
- Cursor Router shows that the model you use depends on the task
- SpaceX is expected to become a historic milestone for AI and space infrastructure
- Impeccable 4 focuses on a “world builder” for creative design
- Grok Build goes open source and the community explodes
- Codex Live Build shows that AI coding remains a hot arena
- An “absurd” question on Grok reveals the appeal of generative AI
- An Ottawa trial and the story of digital footprints on social media
- An AI creation platform is opening interest registration
- 6.2 GW of AI infrastructure in one day: the race has gone far beyond experimentation
- Chance introduces a “trust layer” for AI agents that handle money
- Anthropic releases a free 4-hour course on using Claude like its own team
- A perspective for the Vietnamese market
- References
This week’s AI wave shows the market shifting from hype to real-world deployment: from compute infrastructure and developer tools to applications in advertising, public services, and transaction safety. For Vietnamese marketers, this is not just a technology story, but a signal of how quickly content production, campaign operations, and AI-driven buying, selling, and optimization are changing.
-
Key points:
- Google, OpenAI, Anthropic, and infrastructure companies are pushing AI from the model layer into products, tools, and revenue-generating ecosystems.
- Businesses are beginning to use AI to create ads, websites, automate workflows, and support public services; operational efficiency is a standout theme.
- Alongside the acceleration of commercialization are major questions about model safety, incident transparency, and control over AI spending and decision-making.
- Vietnamese marketers should closely follow the “pragmatic AI” trend: saving costs, speeding up production, and tying everything tightly to performance measurement.
Grok is put on the spot against an AI version of the Odyssey
A satirical post on X pushed Grok into the spotlight by comparing its ability to recreate a classic work like Odyssey with AI. Although the wording was sarcastic, the story behind it points to a real issue: the quality of AI model outputs when handling long-form creative content remains a very demanding test.
For marketers, this is a familiar but still relevant reminder: AI can help sketch ideas, but to create work with emotional depth, brand rhythm, and consistency, human editing is still needed at the end. Source: X.com
A call for more serious investigations into AI misalignment incidents
Ryan Greenblatt said the incident involving OpenAI should be investigated more thoroughly and transparently, especially in cases of concerning “misalignment.” A notable part of his proposal is the list of information that should be disclosed: the prompt, a redacted transcript with sensitive data removed, which model was tested, how it was monitored, and what the model was willing to do to achieve its goal.
This is a very important perspective for the marketing industry as AI is increasingly assigned more automated tasks. If a creative task, customer service workflow, or media optimization process is pushed beyond limits by a model, businesses need clearer testing procedures, behavior logs, and levels of human intervention. Source: X.com
Google reports that AI is driving Search and ad growth
Nick Fox said Alphabet continued to post a strong quarter thanks to its full-stack strategy, with AI playing a central role. According to him, AI features are driving higher Search usage, AI Overviews and AI Mode are contributing to more total queries, and AI Mode has already reached more than 1 billion monthly active users. On the advertising side, more than half of SMB customers have used AI to create or optimize creative.
The notable point is not only the scale of users, but also Google’s emphasis on the infrastructure behind it: handling record query loads and reducing the cost of serving AI Mode. For marketers, the message is clear: AI is no longer just an add-on layer, but is becoming default infrastructure that directly affects search, advertising, and measurement. Source: X.com
Tesla continues tying AI to self-driving cars, robots, and energy
In its earnings call summary, Tesla emphasized that FSD is a significant demand driver in licensed markets, while also saying Cybercab, Tesla Semi, and Optimus are all progressing on their production roadmap. In its shareholder update, the company said Robotaxi has been deployed in 7 major U.S. cities and that Optimus production in Fremont has begun after Model S and X lines were removed.
Although the vision is highly ambitious, Tesla also stressed the need for cautious expansion to avoid accidents or legal backlash. This shows that AI in the real world is not just about models, but also about operational safety, deployment scale, and market trust. Source: X.com, X.com
Anthropic launches an official prompt library for Claude Code
Anthropic has launched an official prompt library for Claude Code, offering ready-made templates for developers to see how they structure context, instructions, and reasoning. This is a notable move because instead of only selling a model, Anthropic is gradually standardizing “how to use the model” as part of the product.
For marketing teams using AI to support coding, automation, or dashboard building, a sample prompt library like this helps shorten trial-and-error time and improve output consistency. Source: X.com
An AI ads platform startup “grew out of” internal needs
TJ Bongiorno said his team is spinning an AI advertising platform out into a separate company after building it to serve internal needs. The platform can create video ads in about a minute using only a prompt, including the angle, copy, and complete creative.
This is a very noteworthy trend for marketers: the most useful AI tools are often not surface-level “wrappers,” but systems that solve a media team’s real pain points. As creative production costs rise, platforms like this could change how ideas are tested, ad variations are created, and launch cycles are shortened. Source: X.com
ServiceNow expands its AI platform across nearly all of the U.S.
ServiceNow said its AI platform has been deployed in nearly all 50 U.S. states, while serving about 90% of the Fortune 500. Examples cited include California using AI to support housing services, Hawaii rolling out a shared platform in six weeks, and Raleigh cutting service desk costs by 66% and saving more than 1,300 work hours per year.
The common thread in these cases is that AI does not replace public agencies, but helps service departments respond faster, with more control, while keeping humans in the loop. This is an important cue for Vietnamese organizations looking to digitize request intake or citizen and customer support workflows. Source: X.com
AMD partners with Anthropic in a major AI infrastructure deal
AMD is said to have announced a strategic deal with Anthropic for 2GW of MI450 GPUs, with the first deployment expected in the first half of 2027. In addition, AMD will invest up to $5 billion in Anthropic.
The market message here is that the AI race is shifting strongly to the infrastructure layer: chips, power, data centers, and large-scale deployment capacity. This indirectly affects marketers because the cost and speed of AI tools depend deeply on the infrastructure game behind them. Source: X.com
Tesla says Cybercab is in production, Robotaxi expands to 7 metros
In its shareholder update, Tesla said Cybercab has begun production at Gigafactory Texas, Semi remains on track for mass production this year, Megafactory Texas is nearly complete, and Optimus will be produced this year. The company also said Robotaxi deployment in the U.S. has expanded to 7 major cities.
Although these milestones still need to be watched as they move from announcements to actual production, they are still a clear signal that Tesla is packaging AI into physical products and operating services. For the marketing industry, “AI + hardware + autonomy” stories often have strong reach because they tap into a very tangible vision of the future. Source: X.com
Moonshot news and the suspicion of Anthropic “distillation”
Ryan Greenblatt repeated an allegation that Moonshot AI “distilled” Anthropic’s Fable model to develop K3, while also describing how the company built an internal platform to flexibly switch between multiple access methods in order to avoid detection. He also argued that Fable’s recent launch does not automatically rule out unauthorized access before release.
This is a sensitive but very important topic in AI: the boundary between legitimate learning, model optimization, and intellectual property theft. For marketers using AI tools, it is a reminder that data provenance, copyright, and vendor transparency will increasingly become product selection criteria. Source: X.com
OpenAI expands hard spend limits to all API accounts
OpenAI Developers said the hard spend limits feature on the API Platform will be opened to all accounts this week. In practical terms, this is a very necessary cost-control tool for product teams and marketing teams using APIs to generate content, personalize, analyze, or run agents.
In an era where AI spends by the minute, the ability to set spending caps helps businesses control budgets and avoid unexpected costs when testing at scale. Source: X.com
ChatGPT Sites: turn ideas into websites right inside the conversation
Some users are highlighting the usefulness of Sites in ChatGPT, a tool that lets you turn an idea into a website or app hosted directly inside ChatGPT. According to the description, users can build static sites or full-stack apps, store data, upload files, add authentication, connect APIs, view analytics, and publish instantly with a shareable URL.
For marketers, this is a very fast way to build landing pages, microsites, interactive tools, lead capture forms, or internal dashboards without waiting for the traditional dev cycle. The speed of idea testing can become a real competitive advantage. Source: X.com
Cursor Router shows that the model you use depends on the task
Cursor launched Cursor Router, an intelligent model router that automatically selects the right model for each task and, according to the company, delivers frontier-quality results at 60% lower cost. This is a clear sign of an emerging trend: you do not always need the biggest model, but the most suitable one.
For marketing teams using AI for content writing, data analysis, or code automation, this kind of routing can significantly reduce costs while keeping output quality good enough. The core lesson is to optimize the system, not just the prompt. Source: X.com
SpaceX is expected to become a historic milestone for AI and space infrastructure
Cathie Wood said SpaceX could become one of the most important companies in global history, not only because of its launch capabilities but also because of its role in satellite telecom networks and orbital data centers. She also said this could be the foundation for building frontier AI models at lower cost.
Although this view is long-term and predictive, it reflects a notable reality: AI is becoming increasingly tied to energy, connectivity, and physical infrastructure. Companies that control both the “pipes” and the models will have a huge advantage. Source: X.com
Impeccable 4 focuses on a “world builder” for creative design
Paul Bakaus introduced Impeccable 4 as a “world builder” tool for new-from-scratch or redesign tasks. According to him, the tool is optimized for frontier models, simpler to use, and especially suited to mobile design workflows, with support for Apple HIG, Material 3, and accessibility checks.
For marketers and designers, this is a clear sign that AI in design is moving from “quick suggestions” to “structured output systems.” That can help shorten the concept and prototype stages. Source: X.com
Grok Build goes open source and the community explodes
Early numbers for Grok Build show the community reacted very quickly after the project went open source: tens of thousands of stars and thousands of forks in a short time. This phenomenon shows that for AI tools aimed at building and creativity, the value of the community ecosystem can be just as large as the model itself.
As marketers need flexible tools to test campaigns, the open-source AI trend helps shorten the time needed to adopt new tools while also allowing technical teams to customize more deeply. Source: X.com
Codex Live Build shows that AI coding remains a hot arena
OpenAI Developers promoted a Codex Live Build session with creators and developers. Even though it was just a livestream, it still reflects the intense competition in AI coding, where tools not only need to write code but also fit closely into real developer workflows.
For marketers using automation, no-code, or low-code, following build sessions like this helps quickly track new tool trends and integration possibilities, making it easier to choose a better stack for the team. Source: X.com
An “absurd” question on Grok reveals the appeal of generative AI
A user tried asking Grok about a scenario in which Chicago is visited by an invincible grizzly bear that only attacks criminals and never leaves the city. The question may seem purely for entertainment, but it actually shows that users see AI as a tool for simulating the world and testing “what if” assumptions.
This is a very familiar type of interaction in creative marketing: content teams use AI to break through idea blocks, test unusual angles, or build concepts before investing in production. Source: X.com
An Ottawa trial and the story of digital footprints on social media
A post said Ezra Levant is being cross-examined in an Ottawa trial related to his social media posts and personal views. Although this is not an AI story in the technological sense, it is still an important reminder about digital traces, public content, and the impact of online speech.
For marketers and personal brand builders, the lesson is that every piece of content on digital platforms can be stored, cited, and scrutinized later. In the age of AI search and summarization, digital footprints have an even longer lifespan. Source: X.com
An AI creation platform is opening interest registration
Nobs said it is continuing to refine its questionnaire to better understand the community’s views before launching an AI content creation platform with its own economic credit mechanism. Although details are still limited, the signal shows that the AI tools market remains open to new product models, especially in creation and usage-resource management.
For Vietnamese businesses, this is a reminder that pricing models, credit mechanisms, and access controls will become just as important as model quality. Source: X.com
6.2 GW of AI infrastructure in one day: the race has gone far beyond experimentation
A roundup on X claimed that three major AI infrastructure announcements in a single day added up to about 6.2 GW. The figure was drawn from information about OpenAI, SpaceXAI, and AMD + Anthropic, with the argument that power capacity is being rapidly redirected toward AI.
Although the individual figures should be read carefully in the context of each original announcement, the broader picture is very clear: AI is entering a phase of industrial-scale consumption of capital, electricity, and compute capacity. This is something marketers need to understand in order to forecast costs, innovation speed, and tool readiness over the next few years. Source: X.com
Chance introduces a “trust layer” for AI agents that handle money
Chance introduced a trust layer for AI agents that can touch money, built around a “propose – verify – execute” mechanism. Every agent action is checked by an independent AI judge against the user’s actual intent before money is transferred, in order to block hallucinations, prompt injection, and out-of-scope transactions.
This is a very important piece of the future of commercial AI. As agents begin handling payments, investments, or ad budget allocation, an independent control layer will become mandatory rather than an add-on feature. Source: X.com
Anthropic releases a free 4-hour course on using Claude like its own team
A free 4-hour course on how to use Claude in the same style as Anthropic’s own workflow is being widely shared by the community. The content focuses on how to prompt correctly, why Claude struggles with code, daily workflows, and a few tweaks that make the model smarter in practice.
For marketers, materials like this are very useful because they turn AI from a “I already know this” tool into a repeatable workflow. Learning how to prompt, organize context, and build workflows will deliver far better results than using AI only at an experimental level. Source: X.com
A perspective for the Vietnamese market
The biggest common thread in this week’s news is that AI is moving from models to operating ecosystems: companies are buying GPUs, setting API cost caps, building internal tools, launching courses, standardizing prompts, and adding control layers before agents touch money or critical workflows. For Vietnamese businesses, this opens up three clear priorities.
First, use AI to speed up production while still keeping human editing at brand touchpoints. Second, invest in workflow infrastructure: prompt libraries, data governance, API costs, behavior logs, and output testing. Third, do not view AI only as a content-writing tool; see it as an operating layer that can affect media buying, lead generation, customer service, and analytics.
In short, the advantage no longer lies in whether you use AI, but in who uses AI more systematically, more safely, and with clearer measurable results.
See more marketing news and guides at https://marketing365.vn.
Follow more updates from Marketing365 to stay on top of the latest marketing trends.
Read more articles in the same category AI developments.
This article focuses on the latest AI news with a perspective for the Vietnamese market.
References
- If Grok made a full length remake of the Odyssey… what are the chances it would be 10X better than that POS…
- The OpenAI incident should be investigated more seriously and more information should be released about what…
- Another strong quarter of Alphabet earnings driven by our full-stack approach!
- $TSLA Q2 2026 Earnings Call Notes (Gali Style)
- Q2 Shareholder Update → Highlights
- ANTHROPIC LAUNCHES OFFICIAL CLAUDE CODE PROMPT LIBRARY
- So, @WiFiMoneyGuy and I accidentally built (and are about to launch) a new business
- JUST IN: $NOW ServiceNow Expands AI Platform Across Nearly All 50 U.S. States
- AMD $AMD announced a strategic partnership with Anthropic for 2GW of its MI450 series GPUs
- We’re expanding access to hard spend limits in the API Platform to all accounts this week
- You should be using Sites in ChatGPT!
- Introducing Cursor Router, our intelligent model router that selects the right model for the task at hand
- ARK Invest CEO @CathieDWood in a new interview on FOX says @SpaceX could become the most important company in
- Introducing Impeccable 4 *world builder*
- Grok Build’s first week as open source, in data
- Watch now: Codex Live Build with @covacut and @davis7
- Hey @grok how would Chicago change if an invincible grizzly bear that feeds on criminal humans,but only stays…
- your interest registration can always be updated until it cannot (closer to launch)
- Three separate AI infrastructure announcements. One single day. 6.2 gigawatts of AI infrastructure. wtf –
- Your AI agent is one hallucination away from draining your account. We make sure that never happens.
- Anthropic just released a FREE 4-hour course on using Claude like its own engineers.
