Building an AI solution goes beyond calling a model. The focus is on creating production-ready AI applications and agents with Microsoft Foundry that can use enterprise data, interact with tools, process different content types, and collaborate to complete real tasks.
| Area | What you should understand |
|---|---|
| Generative AI apps | Build conversational applications using models, APIs, and SDKs |
| Grounding | Connect models to your own data for relevant, fact-based responses |
| Agents + tools | Let agents retrieve information and take actions |
| Multi-agent systems | Orchestrate specialized agents to collaborate on workflows |
| Multimodal AI | Process text, documents, vision, and speech |
| Production | Deploy, publish, monitor, secure, and apply responsible AI safeguards |
Exam focus: Understand not just what these capabilities do, but when and why you would use them together in an Azure AI solution.
A useful mental model for AI-103 is:
User → AI App/Agent → Model → Data + Tools → Action/Response
For more complex solutions:
User → Orchestrator → Agent A + Agent B + Agent C → Tools/Data → Result
An agent therefore isn't simply a chatbot. It combines a model's reasoning capabilities with instructions, knowledge, and tools so it can perform useful work.
The course assumes working knowledge of Python, REST APIs/SDKs, Azure fundamentals, and generative AI concepts. Hands-on practice is important: build applications in Microsoft Foundry, connect models to data, add tools to agents, experiment with multimodal inputs, and create multi-agent workflows.
Key takeaway: Think beyond prompts and models. AI-103 is about assembling the components required for an end-to-end AI solution:
Models → Grounding → Tools → Agents → Orchestration → Production
Generative AI models are powerful, but their trained knowledge is limited. Tools extend models beyond text generation, allowing them to access real-time information, take actions, ground responses in facts, extend functionality, and build intelligent workflows.
| Tool | Purpose |
|---|---|
code_interpreter |
Generate and run code for calculations and data analysis |
web_search |
Find current information on the internet |
file_search |
Search files and ground responses in specific knowledge |
function |
Call custom functions implemented by your application |
Remember: current information → web_search · uploaded/private documents → file_search · calculations/code → code_interpreter · application-specific actions → function
Tools are provided through the tools collection. The model can determine which available tool is appropriate for a request.
response = client.responses.create(
model=model_name,
input="Answer the user's request using the available tools.",
tools=[
{"type": "code_interpreter", "container": {"type": "auto"}},
{"type": "web_search"},
{"type": "file_search", "vector_store_ids": [vector_store.id]}
]
)
print(response.output_text)
Core flow: User → Responses API → Model → Tool → Result → Model → Response
For file_search, documents are stored in a vector store and prepared for semantic retrieval:
Files → Chunking → Embeddings → Vector Store → Retrieval → Model
This lets the model answer using relevant document content rather than relying only on its trained knowledge. Uploaded company policies or private documents → File Search + Vector Store.
Functions are different because the application executes the function, not the model. The model identifies the required function and returns a function-call request:
User → Model → Function Call → Application → Function → Result → Model → Response
The application executes the requested code and returns its result. This process can run in a loop when multiple tool calls are needed.
Key distinction: built-in tools extend the model with predefined capabilities; function calling connects the model to your own application logic and actions.
AI safety is no longer only about what future systems might be capable of. New reports show that people are already trying to use advanced AI for potentially harmful activities.
Anthropic says it has detected and disrupted attempts to misuse its Claude models across several areas, including cyberattacks, surveillance, influence operations and potentially dangerous biological research.
AI can make such activities easier by helping users analyze information, write code, coordinate tasks and automate parts of complex workflows. More capable AI agents could increase this effect by performing multiple steps with less human involvement.
The findings do not mean AI systems are independently launching attacks. They show a different challenge: powerful general-purpose tools can amplify the capabilities of people who misuse them.
For AI providers, businesses and governments, safeguards will increasingly need to combine technical restrictions, monitoring, security testing and human oversight.
APIs and MCP are not competing technologies—they solve different parts of the integration problem.
APIs do the actual work. They let software communicate with services, databases, and other systems. With AI applications, the model itself does not call an API; it chooses an action, while software outside the model executes it.
MCP adds a standardized layer around this process. An MCP server can expose useful actions—such as reading messages or creating tickets—while handling the underlying API calls, authentication, formats, and other implementation details.
This makes integrations easier to discover and reuse across multiple AI applications instead of rebuilding them for each one.
When to use which?
Direct APIs: Simple applications, experiments, or a small number of known operations.
MCP: Multiple AI applications sharing tools and systems.
In short, MCP does not replace APIs. It provides a common, reusable way for AI applications to access the capabilities behind them.
Original video: MCP vs API Explained: Do You Really Need MCP? (en / 17:17) - KodeKloud (YouTube)
larly relevant:
When was the lawn last fertilized? Which month was the hedge trimmed? And what work was done in the garden last autumn? Small details like these are surprisingly easy to forget.
Keeping a record of garden tasks makes it much easier to look back and plan future work. The challenge is finding a structure that is simple enough for everyday use while remaining organized over several years.
A practical approach is to organize the records hierarchically in a note-taking application:
The Garden Chronicle serves as the long-term archive, while each Garden Log contains the records for a particular year.
A simple table works well for the individual entries:
| Date | Task | Notes |
|---|---|---|
| 09 Sep 2026 | Mowed the lawn | Cutting height recorded |
| 15 Sep 2026 | Fertilized the lawn | Autumn fertilizer |
| 03 Oct 2026 | Pruned shrubs | Seasonal pruning |
The Notes column is particularly useful for recording products, quantities, plant varieties, weather conditions, or observations.
With very little effort, this creates a useful garden history that can support future planning and make recurring seasonal tasks easier to track.
Have dozens of tabs open in Brave and want to save them before closing the browser? A simple bookmark export provides an easy backup without installing extensions or running scripts.
First, press Ctrl + Shift + D in Brave. This bookmarks all tabs in the current window and places them together in a folder. Give the folder a recognizable name, such as Open Tabs Backup.
Next:
Open Brave’s Bookmark Manager by entering brave://bookmarks/ in the address bar.
Select the three-dot menu in the upper-right corner.
Choose Export bookmarks.
Select a location and save the resulting .html file.
The exported HTML preserves bookmark titles and URLs, making it useful as a portable backup. It can also be opened in a browser or imported into compatible browsers later.
One limitation is worth knowing: Brave exports the complete bookmark collection, not only the temporary folder containing your open tabs.
Closing a notebook in OneNote does not delete it. Cloud-based notebooks are stored in services such as OneDrive or SharePoint, so permanent deletion must happen at the storage location.
Close it in OneNote. This removes the notebook from the app but leaves its data untouched.
Open OneDrive or SharePoint and locate the notebook in its actual storage location.
Delete the complete notebook rather than individual sections.
Check the recycle bin. A normally deleted notebook remains recoverable until it is removed from the recycle bin or the retention period expires.
Empty the recycle bin if you want to remove it immediately from your accessible storage.
Microsoft 365 business and SharePoint environments may have additional recycle-bin stages or organizational retention policies. These can preserve deleted information even after a user empties the recycle bin.
The key distinction is simple: closing removes a notebook from OneNote; deleting removes it from cloud storage. For permanent removal, always check the underlying storage and its recycle bin.
Europe’s AI ambitions just received a major financial boost. French AI company Mistral has raised €3 billion, giving it a valuation of about €21 billion ($24 billion) and marking the largest equity funding round by a privately owned European technology company.
Mistral develops large AI models and competes in a market dominated by much larger American companies. The new capital is expected to support further model development, computing infrastructure and international expansion.
The investment also has a broader European dimension. Governments and businesses increasingly want greater choice over where their AI technology and data come from. A strong European AI provider could offer another option alongside major US and Chinese platforms.
The enormous investment required to develop advanced AI is concentrating the industry around companies capable of accessing substantial computing power and capital.
Mistral’s latest funding shows that Europe is trying to remain part of that race—not simply as a customer for AI developed elsewhere, but as a producer of its own technology.
The service brings together the key capabilities needed to build and run AI agents without having to manage every technical detail yourself:
With GPT-6 Astra, artificial intelligence is taking another step toward greater autonomy. OpenAI’s new model is designed not only to provide better answers, but also to carry out complex tasks on a computer with less human guidance.
From chatbot to digital worker
Instead of specifying every individual step, users can increasingly define the desired outcome. Astra can then plan and execute multiple steps to reach that goal.
Tasks can include:
More capabilities also mean more risks
Cybersecurity is particularly important. Astra is the first OpenAI model to reach the company’s “Critical” level for cybersecurity capabilities. With suitable tools and permissions, it can potentially discover previously unknown security vulnerabilities. OpenAI has therefore introduced additional safeguards and monitoring systems.
Why it matters
The development highlights a broader shift in AI: from conversational assistants toward systems that can take action. As these systems become more autonomous, clear permissions, strong security controls, and human review become increasingly important.
The AI boom has an unexpected side effect: smartphones and computers are getting more expensive. The reason lies in components found in almost every modern device — memory chips.
AI data centers require enormous amounts of high-performance memory. Manufacturers are therefore dedicating more production capacity to lucrative memory products for servers and AI systems. This leaves less capacity for conventional DRAM and NAND memory used in smartphones, laptops, and SSDs.
The effects are becoming noticeable:
Gartner expects average PC prices to rise by 17% and smartphone prices by 13% in 2026 compared with 2025.
TrendForce forecast further increases in DRAM and NAND contract prices for the third quarter of 2026.
Budget devices are particularly affected because memory represents a larger share of their manufacturing costs.
Manufacturers may respond with higher prices, smaller product ranges, or more conservative memory configurations.
For consumers, this could also change buying habits. Devices may be kept for longer, while used and refurbished smartphones and computers become more attractive.
AI is therefore changing more than software and the workplace. The global infrastructure required to power it is increasingly influencing the price of everyday electronics.
The race to build AI that is both fast and highly capable may be getting more interesting. Google is reportedly preparing Gemini 3.8 Flash, an AI model designed to significantly improve coding performance.
Google’s Flash models are intended to provide a faster, more efficient alternative to its largest AI models. According to reports, Gemini 3.8 Flash has been tested internally with a strong focus on software development.
In Google’s internal coding environment, engineers reportedly preferred the upcoming model over Anthropic’s Opus in some comparisons. However, these are internal evaluations, not independent benchmarks, so real-world performance remains to be verified.
The development points to an important AI trend: smaller, faster models are becoming increasingly capable.
For users, this could eventually mean sophisticated coding assistants and AI agents that respond quickly while requiring fewer computing resources. But until Google officially releases the model and publishes specifications or benchmarks, its exact capabilities remain uncertain.
Artificial intelligence is usually associated with chips, software, and computing power. But behind the AI boom lies an even more fundamental resource: energy. As data centers become larger and more numerous, electricity generation, power grids, and reliable supply are becoming increasingly important.
Expanding digital infrastructure affects an entire supply chain:
Power generation: Data centers require large amounts of electricity, often around the clock.
Power grids: Additional generation capacity has limited value if the grid cannot deliver enough electricity where it is needed.
New energy sources: Alongside renewables, nuclear power is receiving renewed attention, including small modular reactor concepts.
Data centers: Cloud and AI providers increasingly need to consider where sufficient electricity and grid capacity are available.
The key point is simple: regardless of which AI company or chipmaker ultimately succeeds, digital services need energy. Expanding electricity infrastructure could therefore become one of the foundations for further growth in AI and cloud computing.
For businesses and society, this changes the perspective on the AI boom. Progress will not depend solely on better models and faster processors, but increasingly on power plants, electricity grids, energy storage, and available grid capacity.
PVT modules make double use of roof space: they generate electricity while collecting heat for a heat pump. This can be particularly useful where a conventional outdoor heat-pump unit is difficult to install.
A PVT (photovoltaic-thermal) module combines two functions:
Because ambient air is the main heat source, thermal energy can also be collected at night and during winter.
Unlike a typical air-source heat pump, the roof-based collector operates without a fan. Natural airflow provides heat exchange, allowing silent operation without mechanically moving parts on the roof.
PVT can therefore be particularly attractive for terraced houses, densely built areas, or properties with limited space for an outdoor unit.
The required number of modules depends primarily on the building's heating load and system design.
Inflation is falling – yet at the supermarket, in restaurants, or when paying for everyday services, things may not seem much cheaper. How can both be true? The answer lies in what inflation actually measures.
The inflation rate measures how quickly prices are rising on average. If inflation falls from 5% to 2%, goods and services do not automatically become cheaper. They are simply getting more expensive at a slower pace.
A simple example:
Inflation has dropped considerably, but the price is still rising.
For the overall price level to decline, inflation would need to become negative. This is called deflation.
In simple terms:
Individual products can still become cheaper while overall inflation remains positive. Competition, lower production costs, cheaper raw materials, or technological improvements can all push particular prices down.
For households, inflation is only part of the picture. Income growth also matters. If wages rise faster than living costs, purchasing power improves. If they lag behind, everyday life can continue to feel expensive even when inflation has fallen.
In short: As long as inflation remains positive, the overall price level continues to rise. For prices overall to fall, inflation would have to turn negative.
Building more capable AI is increasingly about more than better algorithms. It also requires enormous amounts of computing power — and Anthropic has reportedly secured another major supply.
Anthropic has signed a $35 billion cloud computing deal with AI infrastructure provider Lambda, according to reports published September 1. The agreement involves large-scale computing infrastructure using Nvidia chips.
The deal adds to Anthropic’s already extensive infrastructure commitments. The company has previously announced major capacity agreements involving Amazon, Google, Microsoft and Nvidia as demand for its Claude AI models grows.
Modern AI models require huge clusters of specialized processors both for training and for answering users’ requests. Securing enough chips, electricity and data-center capacity has therefore become a strategic priority for leading AI companies.
For everyday users, these investments may eventually translate into greater AI capacity, faster services and support for increasingly demanding applications. At the same time, the scale of the spending shows how infrastructure-intensive the global AI race has become.
Learning data science becomes much easier when you know where to find good datasets, clear explanations, and real-world examples. These resources cover different stages of that journey.
A good learning strategy is to combine them: learn a concept visually, find a suitable dataset, then build a small project around it. This turns abstract theory into practical experience.
Sometimes the same table or part of a table is needed in multiple places within a Word document. Simply copying and pasting it creates an independent copy: if the original table changes, the copied version remains unchanged. A better approach is to maintain the content once and reference it wherever it is needed.
Within the same Word document, Bookmarks and REF fields provide a practical solution:
TableClasses, and click Add.Ref, then choose the previously created bookmark.Word inserts a reference to the bookmarked content. Changes should be made to the original table. To refresh the references throughout the document, press Ctrl+A → F9.
Important: This works best for a complete table or a continuous range of cells. If you want to display only non-adjacent columns, such as columns 1 and 3 while excluding column 2, a single bookmark is not sufficient. Separate bookmarks or a different table structure may be required.
The Paste Link option is not available for every type of content copied within the same Word document and may therefore appear disabled.
Microsoft Copilot is moving beyond simple chat. New capabilities increasingly connect AI directly with everyday work—helping users see, create, organize, and share information across Microsoft 365.
Copilot Vision lets eligible users share their screen and talk with Copilot about what it sees. It can help examine spreadsheets, explain visible information, or answer questions while you move between applications.
Copilot can transform existing content instead of starting from scratch:
PowerPoint: Create presentations from documents while reusing the style of an existing deck.
Excel: Apply skills such as company branding to selected worksheets.
SharePoint: Turn spreadsheet data into interactive dashboards with filters and summaries.
A centralized Tasks view brings together Copilot activities that are running, completed, or waiting for input, making longer AI-assisted workflows easier to follow.
Copilot responses can be shared through controlled links within an organization. Access can later be revoked when information should no longer be available.
These changes show a clear direction: Copilot is becoming an AI layer across everyday work, rather than simply a chatbot in a separate window.
AI-generated images and videos are becoming convincing enough to blur the line between real events and synthetic content — and today’s news provides a striking example.
U.S. President Donald Trump shared a video appearing to show explosions on Iran’s Kharg Island, an important oil-export hub. However, no independent evidence confirmed an attack on the island at the time of reporting. Analysis using AI-detection software indicated that the footage was probably synthetically generated.
The incident highlights a growing challenge: realistic AI content can spread during fast-moving events before its authenticity is established. When shared by influential accounts, such material can reach large audiences almost instantly and potentially affect public perception.
Before sharing dramatic footage:
Fresh speculation is once again raising questions about whether Grand Theft Auto VI will arrive on time. For now, however, the official release date has not changed.
New reports on August 29 point to online rumors suggesting that development may be less advanced than expected. These claims are linked to reportedly leaked material and have renewed fears of another delay. Importantly, they remain unverified and should not be treated as confirmation of a schedule change.
Rockstar Games continues to list November 19, 2026 as the launch date for GTA 6 on PlayStation 5 and Xbox Series X|S.
Publisher Take-Two also reiterated the November date in its latest financial results earlier this month.
Despite today’s renewed speculation, there is currently no official announcement of another GTA 6 delay. Until Rockstar or Take-Two says otherwise, November 19 remains the date to watch.
One of the most popular ways to use AI for programming is facing a significant change. OpenAI plans to end its agreement with Cursor, the AI coding platform now owned by SpaceX, highlighting how quickly alliances can shift in the competitive AI industry.
Following SpaceX’s acquisition of Cursor developer Anysphere, OpenAI has decided to terminate Cursor’s access under their existing agreement. The service is expected to be fully discontinued on November 12, 2026.
Cursor is not dependent on a single AI provider. Its platform can offer models from several companies, including Anthropic and Google, alongside models associated with its new owner.
The dispute illustrates a broader trend: AI applications increasingly depend on models controlled by competing technology companies.
For users and businesses, this means an AI tool’s capabilities can change when partnerships, ownership or commercial agreements change. Supporting multiple AI models may therefore become increasingly important for services that want to avoid dependence on a single provider.
A theory or learning document can quickly become difficult to scan: technical terms, explanations, examples, additional information, and important rules all look similar. A simple and consistent color system makes these different types of information easier to recognize. The challenge is choosing colors that work both on screen and when printed.
| Content | Formatting | Color | HEX |
|---|---|---|---|
| Technical terms | Bold + color | Dark blue | #1F4E79 |
| Examples | Color | Dark green | #007A3D |
| Rules / key points | Bold, optionally a subtle box | Dark red accent | #9C0006 |
| Additional information / translations | Italic + color | Medium gray | #7F7F7F |
| Body text | Regular | Black | #000000 |
Blue for technical terms and green for examples are not universal standards. What matters most is using each color consistently for the same purpose throughout the document.
Our solar system consists of the Sun, eight planets, and numerous smaller celestial bodies. The planets travel around the Sun in an orbit, while gravity keeps them on their paths.
Example: Earth takes approximately 365 days to complete one orbit around the Sun.
Additional information: Besides the eight planets, the solar system includes dwarf planets, moons, asteroids, and comets.
Remember: A planet does not produce its own visible light; it reflects light from a star.
For individual highlights, Character Styles are particularly useful. Choose Default Paragraph Font as the base and define only the properties that should change, such as the font color. Existing formatting such as bold or italic will then remain intact.
Since colors may disappear when printing in grayscale, important distinctions should also use bold, italic, indentation, or boxes rather than relying on color alone.
The AI boom isn’t happening only online. Artificial intelligence companies are increasingly establishing offices, research centers and other operations across Europe as competition for talent and customers intensifies.
New investment data shows that AI-related foreign direct investment announcements in the EU and UK reached a record 107 projects in the second quarter of 2026. That compares with 82 in the previous quarter and 63 during the same period in 2025.
The expansion illustrates how AI is becoming an important part of the wider economy. Building AI businesses requires more than computing infrastructure—it also requires researchers, engineers, sales teams and specialized services.
For European cities, this could mean new jobs, investment and growing technology ecosystems. At the same time, cities will increasingly compete to attract highly skilled workers and AI companies.
The trend is another sign that the global AI race is moving beyond developing better models. Where AI companies build their teams and operations is becoming important too.
Learning a new language doesn’t have to start with endless vocabulary lists and grammar rules. What matters is using the language as early as possible and focusing your learning on situations you actually encounter in everyday life.
Instead of “I want to speak French,” try: “By Friday, I want to order in French at a café.” Small, practical goals show you what you really need to learn.
Structures like “I would like ___, please,” “Where is ___?” or “Can you help me?” can be reused again and again. A few patterns allow you to express many different things without knowing every grammar rule.
Start with words that appear frequently in everyday life. Don’t learn them in isolation—use them in short sentences and typical situations.
Many languages share words that look or sound similar. Recognizing these related words can make building your vocabulary much easier.
Switch your smartphone or individual apps to your target language. Add TV shows, podcasts, music or social media to your routine. Content you mostly understand but that still contains new words is particularly useful.
Practice typical conversations or describe what you’re doing: “I’m opening the door,” “I’m drinking water.” You’ll quickly discover which words you actually need.
With shadowing, you listen to short recordings and imitate them as closely as possible. Pay attention to pronunciation, rhythm, speed and intonation.
Look for opportunities to speak with others and ask them to correct you. Mistakes provide valuable clues about what to work on next.
With spaced repetition, you review words at increasing intervals. Connect difficult words with images, stories or familiar places to make them easier to remember.
Short daily sessions can be more sustainable than occasional learning marathons. If your progress stalls, change your approach—for example, switch from vocabulary exercises to podcasts or conversations.
Perfection isn’t the goal: Language skills develop through many small successes, practical use and consistent practice.
AI-generated video is becoming more capable — and increasingly useful beyond entertainment. Alibaba Cloud has expanded access to Wan3.0, its latest generative AI model for creating videos from a surprisingly broad range of source material.
Wan3.0 can generate videos of up to 30 seconds and work with references including text, images, audio, existing video, documents and web pages. This means information from presentations, spreadsheets or other documents can potentially be transformed into short visual content instead of requiring a traditional video-production workflow.
The model also combines visual generation with audio and supports resolutions up to 1080p.
The development illustrates a broader shift in generative AI: models are increasingly able to understand several types of media and turn them into finished content. For businesses and creators, tools like this could make product demonstrations, educational clips and social-media content faster to produce.
Human review remains important, however. AI-generated videos can contain inaccuracies, misleading details or material that raises copyright and authenticity concerns.
A horizontal line can make longer notes much easier to scan. However, OneNote does not provide a standard horizontal divider that automatically stretches across the full width of the text area. Fortunately, there is a simple text-based workaround that works well for everyday notes.
Type a row of underscores _ to create the appearance of a continuous horizontal line:
Text above the line
___________________________________________________________________
Text below the lineUnderscores work particularly well because the individual characters appear connected, creating a cleaner line than multiple hyphens (---).
Around 50 to 100 underscores are usually enough. The line does not need to reach the exact right edge of the page—the goal is simply to create a clear visual separation between sections.
Adding an empty line after the divider also creates comfortable spacing between the line and the following content.
You can use OneNote's Draw tools to insert a straight line, but its length is fixed and does not automatically adapt to the width of your content. A table can also be used as a visual separator, although this is unnecessarily cumbersome for such a simple task.
For everyday notes, the underscore method is therefore a quick, clean, and reusable solution.
Important questions, commitments, and deadlines can easily get lost in a busy inbox. This prompt reviews recent emails and turns pending items into a clear, prioritized action list.
Prompt:
Review my emails from the past two weeks and identify all pending tasks, unanswered questions, commitments, and outstanding decisions.
Create a clear table containing:
- topic
- people involved
- short summary
- expected action and possible deadline
- recommendation: Follow up, Take action, Wait, Archive, or Delete
Sort the results by urgency. Combine related emails into one item and avoid duplicates. Highlight unclear cases, missing information, and overdue tasks. Briefly explain each recommendation.
At the end, list the three most important next steps. Do not perform any actions or draft replies without my explicit approval.
Review the suggestions carefully before replying to, forwarding, archiving, or deleting emails. Follow applicable privacy and security policies and treat sensitive information confidentially.