Stories of the week
deepseek.com
DeepSeek Harness launches as open source rival to Claude Code, alongside V4-Pro on API with higher prices
What happened?
DeepSeek, the Chinese AI lab, launched DeepSeek Harness on Thursday, an open-source tool for developers to build and control AI agents. It directly competes with Anthropic's Claude Code. The company also announced the official release of DeepSeek-V4-Pro on its API, but with higher prices than its previous model, DeepSeek-V3.
Why it matters
If you use AI coding assistants, you now have a free and open alternative to Claude Code, which could mean more choices and lower prices down the road. But the price increase for DeepSeek-V4-Pro on the API may raise costs for companies relying on its models. For everyday users, more competition in the AI tools market usually leads to better products and more affordable services.
Impact on your daily life
DeepSeek became known for releasing very powerful and cheap open-source AI models, like DeepSeek-V3. Until now, it focused mainly on models, but with DeepSeek Harness it's also moving into the developer tools used to put AI agents to work. That move directly challenges Anthropic, whose Claude Code is the go-to tool for such tasks. The fact that DeepSeek is raising prices for V4-Pro is notable because it breaks with its earlier strategy of being the low-cost option.
Sources: deepseek.com · venturebeat.com
blog.google
Google unveils Gemini 3.7 Flash AI model for coding, agent workflows
What happened?
On Thursday, August 13, 2026, Google launched Gemini 3.7 Flash, an AI model focused on coding and automated business workflows. The company gave no release date for its flagship Gemini Pro model, leaving speculation about when the bigger version will arrive.
Why it matters
If you rely on AI assistants to write code, fix errors, or automate repetitive workplace tasks, Gemini 3.7 Flash is a new option that promises to do those jobs faster, without waiting for the Pro model. It also matters because Google can integrate it into tools you already use, like Google Cloud, Workspace, and Android, bringing pro-level AI to smaller teams on a budget.
Impact on your daily life
Google has been releasing Flash versions of Gemini as lighter, faster alternatives to its Pro models. With this update, the company is competing directly with OpenAI, Anthropic, and xAI, which are also building specialized AI for coding and autonomous agents. The key twist: Google didn't say when Gemini Pro would launch, leaving businesses unsure whether the most powerful version is weeks or months away.
Sources: blog.google · www.msn.com
z.ai
Zhipu launches flagship model GLM-5.3 as China seeks Mythos-level edge in cyber defence
What happened?
Chinese company Zhipu, also known as Z.ai, launched its flagship model GLM-5.3 and claims it beat leading US systems in cybersecurity tests, including Anthropic's Mythos 5 and OpenAI's GPT-5.6, on the CyberGym benchmark.
Why it matters
This matters because AI-powered cybersecurity affects everyone: banks, hospitals, governments, and the digital services you use every day. If GLM-5.3 really beats Mythos 5 and GPT-5.6 on CyberGym, it means companies won't have to rely only on American models to defend against automated attacks, and China gains ground in a strategic technology that could decide who controls digital security in the future.
Impact on your daily life
Zhipu, known for its GLM family of models, has been positioning itself as the Chinese alternative to OpenAI and Anthropic. In recent years, Beijing has pushed its labs to compete in key areas like cyber defence, and public benchmarks such as CyberGym have become the battleground where models are compared. The release of GLM-5.3 also comes amid ongoing US-China tech tensions, with chip restrictions and AI export controls in the background.
Sources: z.ai · msn.com
openai.com
OpenAI and Anthropic in price war as Chinese AI rivals gain ground
What happened?
OpenAI announced an 80% price cut for GPT-5.6 Luna, its fastest and most affordable model, while Anthropic launched Claude Opus 5, touting 'frontier intelligence' at half the price of Fable 5, its most capable model. The move is a direct response to the growing pressure from Chinese AI rivals.
Why it matters
For you, this means cutting-edge AI tools will become much cheaper. Subscriptions to assistants, developer APIs, and built-in app features could drop in price, making it easier for small businesses and students to adopt AI. However, a price war also brings uncertainty: companies may cut investments in support or safety to keep margins.
Impact on your daily life
The offensive comes as Chinese AI rivals have released models with comparable performance at much lower cost, forcing OpenAI and Anthropic to shift strategy: they are no longer competing only to be the smartest, but to be the cheapest. In the past, model launches focused on capabilities; now price is the main battleground.
Sources: openai.com · anthropic.com · ft.com
What people are talking about
- Artificial (800 mentions)
- Inteligencia (500 mentions)
Skill of the week
github-profile-dashboard (devops-deploy)
Selected for its connection with this week's AI topics.
Tip of the week
practice DeepSeek Harness lanza como rival with your own content: ask it to explain what you don't understand as if you were 12.