by Rhea Purohit in AI & I

TL;DR: Today, we’re releasing a new episode of our podcast AI & I, where Dan Shipper sits down with Andrew Wilkinson, the cofounder of Tiny, a holding company that buys profitable businesses and focuses on holding them for the long term. Watch on X or YouTube, or listen on Spotify or Apple Podcasts.

Plus: We’re hosting an all-day livestream tomorrow with the best vibe coders in the world, showcasing what’s now possible that wasn’t two months ago. Join us on X. On Friday, we’re hosting a free camp for paid subscribers about how agent-native architecture works and how to use it effectively.

---

Entrepreneur Andrew Wilkinson used to sleep nine hours a night. Now he wakes up at 4 a.m. and goes straight to work—because he can’t wait to keep building with Anthropic’s latest model, Opus 4.5

---

Click here to read the full post

Want the full text of all articles in RSS? Become a subscriber, or learn more.

Supporting older operating system versions is often seen as a safe and user friendly choice. The assumption is simple. The more OS versions you support, the more users you can reach. In practice, the cost of backward compatibility is rarely visible at first, but it accumulates steadily and affects development speed, code quality, testing effort, and even product decisions.

This post looks at the real cost of supporting old OS versions and offers practical guidance on how to choose a realistic minimum OS version for a mobile app.

---

Slower Feature Development

Every new OS release brings new APIs, better tooling, and improved platform capabilities. When an app supports old OS versions, developers cannot freely use these improvements.

Instead of implementing a feature once, teams often need to:

• Write conditional code paths for old and new APIs

• Maintain fallback implementations

• Avoid newer platform features entirely

Over time, this leads to a lowest common denominator approach. Features are designed not around what the platform can do today, but around what the oldest supported OS allows. This slows development and limits innovation.

---

Increased Testing Complexity

Supporting a wide OS range multiplies testing effort. Each additional OS version adds more combinations to validate:

• Different system behaviors

• API inconsistencies and edge cases

• Vendor specific issues on older Android devices

Manual testing matrices grow quickly, and automated tests become harder to maintain. Bugs often appear only on specific OS versions that developers no longer use daily, making them harder to reproduce and fix.

The result is more time spent verifying existing behavior instead of building new functionality.

---

Design and UX Limitations

Older OS versions often lack modern UI components, animation capabilities, or system behaviors users now expect. Designers are forced to compromise to ensure consistency across versions.

This can result in:

• Simpler interactions than desired

• Visual inconsistencies between devices

• Features that feel outdated on modern hardware

In some cases, entire UX improvements are postponed or cancelled because they cannot be implemented cleanly on older systems.

---

The Business Impact

From a business perspective, supporting old OS versions is not free. It increases development time, QA costs, and release risk. These costs are often hidden because they appear as slower velocity rather than explicit line items.

At the same time, users on very old OS versions tend to:

• Upgrade less frequently

• Engage less with new features

• Be overrepresented in crash and support reports

This creates a mismatch between effort invested and value returned.

---

How to Choose a Realistic Minimum OS Version

Here are some important factors to consider when deciding on your minimum supported version:

User Distribution

Look at real usage data. If a small percentage of users are on very old OS versions, supporting them may not justify the cost.

Platform Support and Tooling

Align your minimum OS with what the platform vendor actively supports. Tooling, libraries, and documentation are optimized for recent versions.

App Category and Audience

Enterprise apps, internal tools, and professional products can usually move faster than consumer apps with broad audiences.

Feature Roadmap

If upcoming features rely on newer platform capabilities, raising the minimum OS early can simplify delivery and reduce technical debt.

Maintenance Budget

If the team is small, reducing compatibility overhead can significantly improve focus and release quality.

Hello Windows Insiders, today we are releasing Windows 11 Insider Preview Build 26220.7653 (KB5074157) to the Dev Channel which is the same build as the January 16th Windows 11 Insider Preview Build 26220.7653 (Beta Channel) release. If you are an Insider in the Dev Channel, you now have a window to switch from the Dev Channel to the Beta Channel if you would like. This window is soon closing. Important: Once we move Dev Channel forward to a higher build number, the opportunity to switch between these channels will close until a future opportunity becomes available. Changes in Dev Channel builds and updates are documented in two buckets: new features, improvements, and fixes that are being gradually rolled out for Insiders who have turned on the toggle to get the latest updates as they are available (via Settings > Windows Update*) and then new features, improvements, and fixes rolling out to everyone in the Dev Channel. For more information, see the Reminders section at the bottom of this blog post.

Changes and Improvements gradually being rolled out with toggle on*

[Account Settings]

We’re updating the design of dialogs in "Accounts > Other Users" based on modern WinUI framework with support for dark mode: [caption id="attachment_178541" align="alignnone" width="954"] Old 'Add account' dialog user experience for work or school users.[/caption] [caption id="attachment_178543" align="alignnone" width="811"] New 'Add account' dialog user experience for work or school users.[/caption] Please note that the visibility of the above dialog option depends on whether a domain joined work or school account is configured on the device. [caption id="attachment_178539" align="alignnone" width="960"] Old 'Change account type' dialog user experience for other users present on the device.[/caption] [caption id="attachment_178540" align="alignnone" width="814"] New 'Change account type' dialog user experience for other users present on the device.[/caption] Feedback: Please share your feedback in Feedback Hub (WIN + F) under "Settings -> User Profile Account Settings.

[Click to Do]

• The Copilot prompt suggestions in Click to Do will now load instantly. This change is not yet rolling out to Windows Insiders in the EEA (European Economic Area) or China.

[Desktop Background]

• You can now set .webp images for your desktop background in Settings > Personalization > Desktop Background.

Fixes gradually being rolled out with toggle on*

[Taskbar and System Tray]

• Fixed an issue in the latest builds where, when the taskbar was set to autohide, it might invoke before interacting with the bottom of the screen, blocking the use of apps in that area.
• Fixed a recent issue which was causing the Start menu, notification center, and quick settings to not open on click sometimes, although using the keyboard shortcuts for each of these still worked.

[Start menu]

• Fixed a memory leak when opening the Start menu, which could lead to performance issues over time.
• Fixed an issue for Arabic and Hebrew users, where Start menu was opening on the wrong side of the screen when the taskbar icons weren’t centered.
• Fixed an issue where if you opened a folder of pinned apps in Start menu and tried to pan it with touch, it would dismiss Start menu.

[Bluetooth]

• Fixed an issue causing Bluetooth device battery level to not show for some Insiders after the latest builds, even though the devices supported it.

[Xbox full screen experience for PC]

• This build fixes some unexpected app and game behavior when using FSE. We’re continuing to validate scenarios and as such keeping the known issue active for now. Thank you Insiders that have been sharing feedback!

[Login and Lock Screens]

• Fixed an issue which could cause the password icon to appear blank on the login screen.

[Settings]

• Fixed an issue causing Settings to crash when interacting with audio devices in the last two builds.
• Fixed an issue where the keyboard character repeat delay in Settings > Bluetooth & Devices > Keyboard was backwards from how it was set in the backend.

[Other]

• Fixed a few issues with clipped text in msinfo32 when using a high display scaling.
• Fixed an issue impacting some Insiders where clicking on .appinstaller files to install MSIX packages was unexpectedly failing with an invalid argument error.
• Fixed an underlying issue which was causing Notepad to not display an underscore in text with certain fonts when your display was set to 125%.
• Fixed an underlying issue which could cause explorer.exe to hang when first logging into your PC, if certain apps were configured as startup apps. This could make the taskbar not appear.

Known issues

[Xbox full screen experience for PC]

• Some apps may behave unexpectedly when using FSE, particularly those that expect to be locked to a given size or launch additional windows.

[Taskbar & System Tray]

• We’re investigating an issue where for some Insiders apps aren’t showing in the system tray when they should be.

[Display & Graphics]

• [NEW] We’re working on the fix for an issue causing secondary monitors to not display correctly and show black screens after the latest updates for a small number of Insiders.

[Click to Do]

• We're working on the fix for an issue where the Microsoft 365 Copilot prompt box on selected images does not function if the Microsoft 365 Copilot app is not running.

Reminders for Windows Insiders in the Dev Channel

• Updates are based on Windows 11, version 25H2 via an enablement package (Build 26220.xxxx).
• Many features are rolled out using Controlled Feature Rollout technology, starting with a subset of Insiders and ramping up over time as we monitor feedback to see how they land before pushing them out to everyone in this channel.
• For Windows Insiders who want to be the first to get features gradually rolled out to you, you can turn ON the toggle to get the latest updates as they are available via Settings > Windows Update*. Over time, we will increase the rollouts of features to everyone with the toggle turned on. Should you keep this toggle off, new features will gradually be rolled out to your PC over time once they are ready.
• Features and experiences included in these builds may never get released as we try out different concepts and get feedback. Features may change over time, be removed, or replaced and never get released beyond Windows Insiders. Some of these features and experiences could show up in future Windows releases when they’re ready.
• Some features in active development we preview with Windows Insiders may not be fully localized and localization will happen over time as features are finalized. As you see issues with localization in your language, please report those issues to us via Feedback Hub.
• Check out Flight Hub for a complete look at what build is in which Insider channel.

Thanks, Windows Insider Program Team

Build n8n workflows for semantic and hybrid vector search, powered by Oracle AI Database 26ai

Following the announcement of Oracle Database support in n8n, I went about running the sample n8n workflow for Oracle AI Database and hybrid vector search submitted by my colleague, Sudarshan Soma.

This n8n workflow implements semantic and hybrid keyword search on a set of sample recruiting documents with candidate details, and highlights the vector search capabilities of Oracle AI Database 26ai. It identifies the closest document content that matches the search string input.

This blog post walks through the nuts and bolts required to run this n8n workflow using Oracle Database.

What do I need to get started?

• Have a running Oracle AI Database 26ai instance with DBMS_HYBRID_VECTOR PL/SQL package available.
• Install Node.js, npm and the latest n8n version (1.117.0 or newer).
• Download the ONNX LLM model — all_MiniLM_L12_v2_augmented.zip , extract the files and store them in a suitable directory, say /opt/oracle/onnx that is accessible by the Oracle Database instance.

Connect to the Oracle Database instance as a SYSDBA user (like SYS or SYSTEM) using another application like SQL*Plus and create a DBA user, such as, for example,app_admin:

-- Create DBA user
CREATE USER app_admin IDENTIFIED BY <password>
  DEFAULT TABLESPACE users 
  TEMPORARY TABLESPACE temp 
  QUOTA UNLIMITED ON users;


-- Grant privileges
GRANT DBA TO app_admin;
GRANT CREATE TABLESPACE, ALTER TABLESPACE, DROP TABLESPACE TO app_admin;

This DBA user will be used in the n8n workflow initially to setup the data tables, and to run the ONNX LLM model for the semantic and hybrid search.

Let‘s get going then…

Start your n8n server (n8n start). Once the n8n server has started, open the default page http://localhost:5678 in your browser, and set some credentials to login to the n8n server.

Click on the ‘Create Workflow’ button on the top right corner in the n8n homepage. A new workflow canvas then appears:

Now open another tab in the browser, and open the AI-powered document search with Oracle and ONNX embeddings for recruiting workflow.

Click on the ‘Use for free’ button.

Click on the ‘Copy template to clipboard (JSON)’ option to copy the JSON file of this workflow to your local machine’s clipboard.

Now, go back to your n8n page with empty canvas and paste this (use the appropriate paste command for your machine).

We now have the full workflow in our n8n canvas.

There are 4 major tasks in this workflow here. Let’s go through them one-by-one to understand the flow, and configure them accordingly.

Create a local user for the vector operations

The 1st sub-task is to create a local user (say docuser ) for hybrid and semantic vector-based search operations.

To create the local user, you have to use the DBA user ( app_admin ) we created earlier. Double click on the Create a user node first.

Note the docuser credentials set in this node.

Also, set the DBA User ( app_admin ) as the Oracle Credentials Account in the Credential to connect with field, then save it.

Once this node is updated. move on to the ONNX Model Directory Path node.

In this node, set the path, where the ONNX LLM model files are located. In this case, the location is /opt/oracle/onnx .

Finally, we grant the required privileges to the docuser user to access the vector operations and the ONNX LLM model.

Setup the data and vector indexes

The next task is to setup the table, data, ONNX LLM models, and the Vector Index, in order to run the semantic and hybrid searches on the recruiting documents.

Initially, we need to create the table and load the ONNX LLM model in the docuser schema. So, we need to configure the 1st node in the Data and Vector Index Setup section as follows:

Create a new Credentials object by selecting the Create new credential option in the Credential to connect with field to add the details for the docuser user.

Add the docuser credential details in this new Credentials object and save it. Ensure that the database connection goes through successfully.

Ensure that the docuser credentials object is selected for this node.

This node reloads the ONNX LLM model named doc_model, creates the table doc_tab for storing the document contents, clears old table data and drops old hybrid vector indices, if any.

The next node generates the data for the recruiting documents.

The next node adds the document data into the doc_tab table. Remember that this node should use the docuser credentials.

The last node in this task recreates the hybrid vector index for the upcoming vector searches. Note that this node also uses the docuser credentials.

In this case, an index named my_hybrid_idx is created on the text column of the doc_tab table. It uses the ONNX LLM modeldoc_model created earlier.

With that, the data setup for the vector searches is complete.

Run Pure semantic vector search in document mode

The next task runs the pure semantic search on the documents to find the closest match for a predetermined text input.

In this case, we plan to run semantic search on the pre-determined text “prioritize teamwork and leadership experience”.

This text is used as the search text among the recruiting documents. Note that the query below will return the top 3 documents (or candidate details specifically), who fit the requirement of the text “prioritize teamwork and leadership experience”.

Note again that we still use the docuser credentials.

Let’s run the Semantic Search workflow now.

Note that the successful workflow run returns the information of the top 3 candidates that satisfy the criteria “prioritize teamwork and leadership experience”.

Run Hybrid Vector Search based on user input from a chat interface

Now, we can also run vector searches based on user input from a chat interface.

Note that the ‘Perform Hybrid Search Node’ uses the docuser credentials.

If the user wants to find a ‘DevOps Engineer’, they can type it in the chat interface below and get the required candidate details.

Note that the chat window will return the candidate who matches the “DevOps Engineer” requirements.

Any other keyword in the chat interface will return the candidate that has the highest vector score with that keyword. For instance, typing “C Python and Database” would most probably return Candidate 5.

Conclusion

We hope that this blog gives a good starting point into building more robust AI powered workflows with Oracle AI Database and n8n. Let us know your results and happy AI-ing!

References

• AI-powered document search with Oracle and ONNX embeddings for recruiting | n8n workflow template
• Oracle AI Vector Search User's Guide

---

Unlocking Intelligent Search: Vector Search Workflows with n8n and Oracle AI Database 26ai was originally published in Oracle Developers on Medium, where people are continuing the conversation by highlighting and responding to this story.

In this episode, Michael, Sarah and Mark talk to Nick Wryter about agentic AI identity, with a big focus on least privilege issues. We also cover news about:

• Microsoft AI Tour
• Azure Database for PostgresSQL
• Azure MCP Server for Azure Confidential Ledger
• Application Gateway, FIPS 140-2 and TLS
• Outbound internet access from VMs
• Azure NetApp Files and Ransomware protection
• Azure Cosmos DB Mirroring

https://aka.ms/azsecpod

Download audio: https://content.rss.com/episodes/8411/2475840/azsecpodcast/2026_01_21_20_58_02_2dc09243-3f64-459d-9cfe-989d55c052cf.mp3

Anthropic is overhauling Claude's so-called "soul doc."

The new missive is a 57-page document titled "Claude's Constitution," which details "Anthropic's intentions for the model's values and behavior," aimed not at outside readers but the model itself. The document is designed to spell out Claude's "ethical character" and "core identity," including how it should balance conflicting values and high-stakes situations.

Where the previous constitution, published in May 2023, was largely a list of guidelines, Anthropic now says it's important for AI models to "understand why we want them to behave in certain ways rather than just specifying w …

Read the full story at The Verge.