Nội dung
- Grok 4.5 takes the lead on long-horizon benchmarks
- Musk continues to emphasize the “removing censorship” narrative at Twitter/X
- GPT-5.6 is now officially available on Amazon Bedrock
- Rumors over McConnell “proof of life” photo stir social media
- The stock market reacts to AI, semiconductors and geopolitics
- Sam Altman is surprised by Claude’s design capabilities
- Altman says OpenAI’s models are getting better at design
- U.S. open-source AI is accelerating on Ollama
- ChatGPT Work emerges as a mobile “office assistant”
- ChatGPT Sites opens public beta, turning prompts into shareable products
- NVIDIA: the real battlefield is cost per token
- An event poster is suspected to be AI-generated
- Altman continues to talk about the value of user experience
- iOS 27 public beta hints at Siri AI and new features
- Ollama brings open models to Yahoo Finance
- India puts responsible AI into its campaign for a UN Security Council seat
- SpaceXAI says it will delete all previously uploaded user data
- SpaceXAI clarifies data retention policy and the /privacy command
- GPT-5.6 is being used for autoresearch, reducing follow-up questions and staying more focused
- A tool that converts “AI slop” into pixel art is released for free
- A perspective for the Vietnamese market
- References
The AI race is entering a more clearly monetized phase than ever before: from long-horizon benchmarks and inference infrastructure to work tools for enterprises. For Vietnamese marketers, this is not just a technology story, but also a signal about content production speed, cost optimization, and how AI platforms will enter everyday workflows.
- Key points:
- Elon Musk said Grok 4.5 has reached No. 1 on Long-Horizon Terminal-Bench, highlighting the long-task capabilities of AI agent.
- OpenAI, AWS and NVIDIA are all intensifying the enterprise AI race: GPT-5.6 is on Bedrock, Work/Sites are expanding, and inference infrastructure is shifting its focus to cost per token.
- Apple, TSMC and chip suppliers continue to benefit from AI demand, while open-source models such as Nemotron Ultra are growing faster.
- Messages from Sam Altman, Elon Musk and the tech industry show that AI is moving from “demo” to a real-world tool for design, research and operations.
Grok 4.5 takes the lead on long-horizon benchmarks
According to Elon Musk’s post, Grok 4.5 has reached No. 1 on Long-Horizon Terminal-Bench. This is a notable benchmark because it measures an AI’s ability to maintain a plan and complete a long chain of tasks, rather than simply answering briefly or handling steps one by one.
A quoted note from @tetsuoai shows that this progress is significant: the original paper had recorded fairly low limits for the models tested, while Grok 4.5 raised the number of completed tasks to 13 out of 46. What matters for businesses is that model performance is not only about being “smarter,” but also about reducing the number of steps and lowering the cost of repeated context.
Source: Elon Musk (@elonmusk), X
Musk continues to emphasize the “removing censorship” narrative at Twitter/X
In another post, Elon Musk said the “woke mob” had turned Twitter into “Wormtongue to the World,” continuing a familiar argument: that his purchase of Twitter was a turning point in restoring free speech. While this is Musk’s personal view, it shows that X is still being positioned as a platform for public debate rather than a purely entertainment-focused social network.
For marketers, the key point is that the content distribution ecosystem is shifting toward less dependence on a single “safe” tone. At the same time, the risks of media crises, fact-checking and brand trust management are also increasing.
Source: Elon Musk (@elonmusk), X
GPT-5.6 is now officially available on Amazon Bedrock
Greg Brockman said GPT-5.6 has been made broadly available on Amazon Bedrock. According to the AWS quote, the GPT-5.6 Sol, Terra and Luna models are deployed as three layers of intelligence, from advanced reasoning to fast reasoning, running on AWS’s next-generation inference infrastructure.
This is an important step for businesses because it means AI is no longer confined to a chat interface, but goes straight into the cloud architecture familiar to engineering and operations teams. For marketers, this opens up use cases such as data analysis, report generation, content personalization and internal workflow automation at a larger scale.
Source: Greg Brockman (@gdb); AWS Newsroom
Rumors over McConnell “proof of life” photo stir social media
Another thread of discussion, not directly about AI, shows how the digital information environment is becoming easier to question and manipulate. The post claimed that the McConnell hospital photo may have been an old image, and Senator Ron Johnson’s own remarks helped the rumor spread further.
For marketers, the lesson here is that content verification is becoming increasingly important. When images, videos and documents can be called into question, brands need stronger verification processes, faster responses and better trust management than ever before.
Source: Occupy Democrats (@OccupyDemocrats), X
The stock market reacts to AI, semiconductors and geopolitics
The market recap shows multiple forces at work at once: Trump’s statement on the Strait of Hormuz, Apple being mentioned with a new chip strategy for AI servers, TSMC reporting record revenue thanks to AI demand, and Meta continuing to expand data centers.
The takeaway is that AI is no longer a standalone “technology story” but a variable affecting supply chains, compute energy and enterprise value. B2B brands, especially in tech, need to closely monitor infrastructure costs because they directly affect the ability to scale AI products.
Source: amit (@amitisinvesting), X
Sam Altman is surprised by Claude’s design capabilities
Sam Altman wrote that he initially thought something was “satire,” because he was still checking whether the account name had been misspelled as “c1audeai” or something similar. The context was a response related to Claude, showing that AI leaders themselves are closely tracking the output quality of competitors.
The underlying message is quite clear: the model’s ability to design, present and create aesthetic experiences has advanced a great deal. For marketing teams, this is a sign that AI is no longer just generating text, but can also take part in visual creation, layout and concept development.
Source: Sam Altman (@sama), X
Altman says OpenAI’s models are getting better at design
In another post, Altman admitted that seeing OpenAI models be “good at design” still surprises him. This is confirmation from an insider that AI is moving beyond the “writing assistant” stage and deeper into tasks that require aesthetics and product thinking.
For Vietnamese marketers, this is especially useful for banners, landing pages, presentations and rapid concept testing. However, the real value still lies in human editing: choosing the right goal, the right tone and the right brand context.
Source: Sam Altman (@sama), X
U.S. open-source AI is accelerating on Ollama
Ollama said U.S. open-source models are quickly gaining market share, with NVIDIA’s Nemotron Ultra growing well on the platform and supporting longer, more complex tasks for developers.
The open-source trend is often attractive to small and medium-sized businesses because it is flexible, easy to test and can offer better control over internal deployment. For the Vietnamese market, this is an opportunity for product and marketing teams that want to experiment with AI without relying entirely on a single proprietary provider.
Source: ollama (@ollama), X
ChatGPT Work emerges as a mobile “office assistant”
Greg Brockman shared that ChatGPT Work is more noteworthy than it gets credit for, especially when used on a phone. The accompanying quote describes it in a very practical way: if you need answers, use chat; if you need code, use Codex; and if you need planning, research or review, use Work.
This is very close to the needs of modern marketing teams: drafting campaign outlines, reading PDFs, summarizing insights and checking content before publication. AI is shifting from a “content creation tool” to a “work orchestration tool.”
Source: Greg Brockman (@gdb), X
ChatGPT Sites opens public beta, turning prompts into shareable products
ChatGPT Sites is described as a more interesting way to communicate internally within a company. According to the quote, the feature lets users turn prompts, files or rough ideas into dashboards, trackers, reports, prototypes or lightweight apps; users can build, edit and test in Work or Codex, then publish via URL.
This is a very notable direction for marketing teams because it dramatically shortens the path from idea to usable tool. Instead of waiting for engineering to build a prototype, marketers can quickly create internal reports, campaign hubs or progress-tracking dashboards.
Source: Greg Brockman (@gdb); quote from @jxnlco, X
NVIDIA: the real battlefield is cost per token
NVIDIA says that as businesses move from experimentation to real deployment, the infrastructure question has shifted from “peak chip” to “cost per token” — meaning how much useful output each dollar spent can produce, within acceptable latency and power consumption.
This is a key point for any team considering large-scale AI adoption. If operating costs do not come down, AI will struggle to move beyond the experimental stage. As a result, both hardware and inference software become core competitive factors.
Source: NVIDIA (@nvidia), X
An event poster is suspected to be AI-generated
A viral post claimed that the “sumbakkokjil” poster may be AI-generated, and the lack of an official announcement only deepened public suspicion. This is a classic example of the new information environment: a single well-made image can be enough to cause confusion if context is missing.
For communications professionals, this is a reminder to proactively disclose image sources, release schedules and verification evidence. Transparency not only helps reduce crises, but also makes the brand more trustworthy.
Source: haru✶⋆.˚ (@harumanyogi183), X; quote from @256895u
Altman continues to talk about the value of user experience
“Come for the best model, stay because we don’t treat you with contempt” reflects how OpenAI is positioning itself: not just selling model capability, but also the experience of working with users.
For marketers, this is a powerful clue. In the AI era, competitive advantage no longer lies in a single technical metric, but in the entire product experience: speed, kindness, contextual understanding and long-term trust.
Source: Sam Altman (@sama), X
iOS 27 public beta hints at Siri AI and new features
AppleTrack said the iOS 27 public beta has been released, with early access to Siri AI, child safety features and performance improvements. While the specific details still need further verification from Apple, the information suggests Apple is continuing to push AI into its mobile ecosystem.
This is a major signal for mobile marketing, apps and consumer services: as AI assistants appear more deeply inside the iPhone experience, search behavior, interaction and customer support could change significantly.
Source: AppleTrack (@appltrack), X
Ollama brings open models to Yahoo Finance
Ollama said a company co-founder/representative is appearing on Yahoo Finance to discuss open models. The fact that open models are being brought onto a mainstream financial channel shows that the AI story is no longer limited to labs or open-source communities.
For the market, this means open-source AI is entering the conversation around business, revenue and commercialization. Vietnamese businesses should keep a close eye on this, because open models could be a cost-saving path for testing many marketing and product use cases.
Source: ollama (@ollama), X
India puts responsible AI into its campaign for a UN Security Council seat
When launching India’s campaign for the 2028–29 UN Security Council term, Foreign Minister S. Jaishankar clearly outlined a “human-centered” vision for AI, placing it within an inclusive, safe framework that serves the public interest. He also emphasized preventing AI misuse and threats to international peace and security.
This is notable because AI has become a diplomatic issue, not just a technology issue. Businesses deploying AI in Vietnam will soon have to face clearer requirements around ethics, data safety and social responsibility.
Source: Dr. S. Jaishankar (@DrSJaishankar), X
SpaceXAI says it will delete all previously uploaded user data
Elon Musk confirmed as a precautionary measure that all user data previously uploaded to SpaceXAI will be completely deleted. This is a very strong privacy statement, especially as AI products increasingly depend on data to improve quality.
For marketers and businesses, the core question is not only “what can AI do,” but also “is the data retained, used for what, and how can it be deleted?” This is the key to choosing an AI provider that fits internal compliance standards.
Source: Elon Musk (@elonmusk), X
SpaceXAI clarifies data retention policy and the /privacy command
In another post, Musk said SpaceX’s data retention policy is useful for debugging, but still respects privacy. The quote from SpaceXAI states clearly that for teams using zero data retention, no trace or code data is kept; if ZDR is off, users can use the /privacy command to change settings and delete synced data.
The message here is that enterprise AI tools increasingly need to provide clear data configuration options. This has direct implications for agencies and brand teams handling sensitive documents, product strategy or customer data.
Source: Elon Musk (@elonmusk), X; quote from @SpaceXAI
GPT-5.6 is being used for autoresearch, reducing follow-up questions and staying more focused
alphaXiv said they are using GPT-5.6 Sol to reproduce the main results of a study on how knowledge can be memorized but not generalize well in fine-tuning. Compared with GPT-5.5 and even Fable 5, GPT-5.6 is described as being more focused on the important experiments, less distracted by side details and better at handling ambiguity on its own.
This is a highly valuable signal for research workflows, market analysis and hypothesis testing. For marketers working on reports, insights or content strategy, a model that goes straight to the important question can save a great deal of time.
Source: alphaXiv (@askalphaxiv), X
A tool that converts “AI slop” into pixel art is released for free
Finally, an independent developer said a tool that converts “AI slop” images into pixel art has been released, free, with no sign-up required and open source. The creator claims the tool outperforms other downscaling methods in accuracy and detail.
From a content marketing perspective, this is an interesting example of “narrow AI that is genuinely useful.” Not every tool needs a giant model; sometimes specialized utilities deliver clearer value for designers, social teams or creators.
Source: Astropulse (@RealAstropulse), X
A perspective for the Vietnamese market
Overall, these 20 news items show that AI is moving in three very clear directions: stronger long-task capability, lower deployment cost, and deeper integration into real workflows. For Vietnamese businesses, this is the time to move from intuitive experimentation to choosing use cases that can be measured: content editing support, customer research, report summarization, prototype creation and internal automation.
From a marketing perspective, teams should prioritize four things: standardizing verification processes for AI-generated content; building clear data-use and privacy frameworks; tracking cost per task rather than only looking at “model strength”; and training teams to work with AI as a collaborator, not a complete replacement. This is how to benefit from the AI wave while still preserving quality, trust and business efficiency.
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 latest AI news with a perspective for the Vietnamese market.
References
- Grok 4.5 reaches #1 position on Long-Horizon Terminal-Bench — Elon Musk (@elonmusk)
- The woke mob had turned Twitter into Wormtongue to the World — Elon Musk (@elonmusk)
- GPT-5.6 is now generally available on Bedrock — Greg Brockman (@gdb)
- BREAKING: HOLY SH*T! MAGA Senator Ron Johnson reveals that the Mitch McConnell proof of life photo was likely… — Occupy Democrats (@OccupyDemocrats)
- A TON OF THINGS HAPPENED IN THE STOCK MARKET TODAY. Here’s a full recap: — amit (@amitisinvesting)
- i thought this was satire, kept looking for the handle to be spelled c1audeai or something — Sam Altman (@sama)
- still sorta breaks my brain to see our models be good at design finally — Sam Altman (@sama)
- U.S. open-source models are quickly gaining ground. @Nvidia’s newest Nemotron Ultra is fast growing on Ollama… — ollama (@ollama)
- ChatGPT Work has gotten less attention, but it’s really cool. Love using it from my phone: — Greg Brockman (@gdb)
- Try out ChatGPT Sites, makes it way more fun to communicate things within a company: — Greg Brockman (@gdb)
- As organizations move from AI pilots to production, infrastructure decisions have shifted from peak chip… — NVIDIA (@nvidia)
- damn it so sumbakkokjil poster was AI?? it was all fake?? and they never even announced cortis was going to… — haru✶⋆.˚ (@harumanyogi183)
- come for the best model, stay because we don’t treat you with contempt — Sam Altman (@sama)
- The iOS 27 public beta has just been released. You can sign-up now for early access to Siri AI, new child… — AppleTrack (@appltrack)
- Ollama’s @jmorgan is live on Yahoo Finance to talk about open models. — ollama (@ollama)
- Pleased to launch India’s campaign for the @UN Security Council 2028-29. Underlined 🇮🇳’s approach rooted in… — Dr. S. Jaishankar (@DrSJaishankar)
- True. As a precautionary measure, all user data that was uploaded to SpaceXAI before now will be completely… — Elon Musk (@elonmusk)
- SpaceX policy regarding data retention. It is actually helpful for debugging issues if we can retain some… — Elon Musk (@elonmusk)
- Introducing autoresearch with GPT 5.6 — alphaXiv (@askalphaxiv)
- The AI slop -> pixel art converter I’ve been building is released! — Astropulse (@RealAstropulse)
