---

Welcome to episode 335 of The Cloud Pod, where the forecast is always cloudy! Welcome to the first show of 2026, and it’s a full house, too! Justin, Jonathan, Ryan,  and Matt are all here to reflect on 2025, plus bring you their predictions for 2026.

Let’s get started! 

Titles we almost went with this week

• SQL Me Maybe: AlloyDB Gets Chatty With Your Database **OpenAI
• SELECT * FROM natural_language WHERE accuracy LIKE ‘100%’ **Anthropic
• etcd You Were Worried About Database Limits: CloudWatch Has Your Back
• CSV You Later: Looker Adds Drag-and-Drop Data Uploads
• AWS Spots an Opportunity to Manage Your Container Costs
• EKS Network Policies: No More IP Address Whack-a-Mole
• AWS Security Hub Splits: It’s Not You, It’s CSPM
• Spot On: ECS Finally Manages Your Cheapest Compute
• TOON Squad: DigitalOcean’s New Format Makes JSON Look Bloated
• The Price is Wrong: AWS Breaks Two Decades of Downward Pricing Tradition
• Show Your Work: Why AI-Generated Code Without Tests is Just Expensive Spam
• No More Agent Orange: Google Simplifies VM Extension Deployment
• AWS Discovers Prices Can Go Both Ways, Raises GPU Costs 15 Percent
• Sovereignty Washing: When Your European Cloud Still Answers to Uncle Sam
• Agent Builder Gets a Memory Upgrade: Google’s AI Finally Remembers Where It Put Its Keys
• Ctrl+F for the Future: A year-end Scorecard & Next-Gen Bets
• AI Agents, GPU Prices, and The best of the Cloud Pod 2025
• Beyond the Hype: The Cloud Pods Definitive 2025 Year in Review
• Apocalypse Now… What? Our 2026 Forecast

 

Follow Up 

01:27 RYAN’S PREDICTIONS

Prediction Status Notes Quick LLM models for individuals ACCURATE Meta-Llama-3.1-8B-Instruct, GLM-4-9B-0414, and Qwen2.5-VL-7B-Instruct—each chosen for an outstanding balance of performance and computational efficiency, making them ideal for edge AI deployment. A new AI inference application called Inferencer allows even modest Apple Mac computers to run the largest open-source LLMs. AI at the edge natively (Lambda-esque) ACCURATE Akamai launched a new Inference Cloud product for edge AI using Nvidia’s Blackwell 6000 GPUs in 17 cities. AWS IoT Greengrass with Lambda functions for edge logic. “Edge AI allows for instant decision-making where it matters most—close to the data source.” Cloud native security mesh multi-cloud UNCLEAR Service mesh technologies continue to evolve (Istio, Linkerd), but I didn’t find a breakthrough “app-to-app at the edge” security mesh product announcement in 2025. This one needs more specific evidence.

Ryan Score: 2/3

02:25 MATTHEW’S PREDICTIONS

Prediction Status Notes FOCUS adopted by Snowflake or Databricks ACCURATE FOCUS version 1.2 was ratified on May 29, 2025. Three new providers announced support: Alibaba Cloud, Databricks, and Grafana. Databricks officially adopted FOCUS! AI security/ethical standard (SOC or ISO) ACCURATE ISO 42001 is the first international standard outlining requirements for AI governance. Major companies achieving certification in 2025: Automation Anywhere is among the first 100 companies worldwide to earn ISO/IEC 42001:2023 certification. Anthropic also achieved ISO 42001 certification. Amazon deprecates 5+ services (WorkMail bonus) ACCURATE (no bonus) 19 services are mothballed, four are being sunset, and one is end of its supported life. Deprecated services include CodeCommit, Cloud9, S3 Select, CloudSearch, SimpleDB, Forecast, Data Pipeline, QLDB, Snowball Edge, and more. WorkMail NOT deprecated – WorkDocs was (April 2025), but WorkMail remains active.

Matthew Score: 3/3

03:22 JONATHAN’S PREDICTIONS

Prediction Status Notes Company claims AGI achieved ACCURATE Integral AI, founded by ex-Google veteran Jad Tarifi, claims to have built a world-first AGI model (December 2025). Also, Sam Altman called GPT-5 “a significant step along the path to AGI” at release. AI agents booking reservations/real-world tasks FULLY ACCURATE OpenAI’s Operator can execute tasks like filling out forms, managing online reservations, and even booking tickets to sporting events. Google AI Mode’s agentic capabilities help take the hassle out of booking restaurant reservations, event tickets, or beauty and wellness appointments. Models that can learn in real-time PARTIALLY ACCURATE Extended context windows and memory systems have improved dramatically. Claude 4 has “memory capabilities, extracting and saving key facts to maintain continuity.” However, true real-time learning/weight updates during conversations haven’t fully materialized yet.

Jonathan Score: 2.5/3

05:07 JUSTIN’S PREDICTIONS

Prediction Status Notes GPT-5, Claude 4, and Gemini 3.0 FULLY ACCURATE GPT-5 (August 7, 2025), Claude 4 (May 22, 2025), Gemini 3 (November 18, 2025). All three major models have been released! Plus, we’ve already seen GPT-5.1, GPT-5.2, and Claude Opus 4.5. OpenAI is not seen as a leader ACCURATE ChatGPT’s user growth is slowing, and Google’s Gemini is gaining ground. Anthropic now holds 32% of the enterprise LLM market share by usage, with OpenAI at 25%—a sharp reversal from 50% vs. 12% in 2023. Sam Altman issued a “code red” memo following the release of Gemini 3. 10+ companies RTO 5 days after Q2 PARTIALLY ACCURATE Major announcements after Q2: Novo Nordisk, Paramount Skydance, NBCUniversal, Instagram, Starbucks, Samsung, Freddie Mac. Many 5-day mandates took effect in 2025 (Amazon, AT&T, JPMorgan, Dell), but several were announced pre-Q2. Close call.

Justin Score: 2.5/3

JONATHAN’S PREDICTIONS

Prediction Status Notes Company claims AGI achieved ACCURATE Integral AI, founded by ex-Google veteran Jad Tarifi, claims to have built a world-first AGI model (December 2025). Also, Sam Altman called GPT-5 “a significant step along the path to AGI” at release. AI agents booking reservations/real-world tasks FULLY ACCURATE OpenAI’s Operator can execute tasks like filling out forms, managing online reservations, and even booking tickets to sporting events. Google AI Mode’s agentic capabilities help take the hassle out of booking restaurant reservations, event tickets, or beauty and wellness appointments. Models that can learn in real-time PARTIALLY ACCURATE Extended context windows and memory systems have improved dramatically. Claude 4 has “memory capabilities, extracting and saving key facts to maintain continuity.” However, true real-time learning/weight updates during conversations haven’t fully materialized yet.

Jonathan Score: 2.5/3

FINAL STANDINGS

Host Score Grade Matthew 3/3 A+ Justin 2.5/3 A Jonathan 2.5/3 A Ryan 2/3 B+

Key Takeaways for the Pod

1. The AI model predictions were NAILED – All three major model releases happened exactly as predicted.
2. OpenAI’s dominance really did slip – Anthropic now leads enterprise, Gemini is surging, Sam issued “code red.”
3. AI agents are HERE – OpenAI Operator and Google AI Mode are booking real reservations.
4. AWS deprecation wave was massive – Way more than 5 services axed (but WorkMail survived!)
5. Edge AI exploded – Akamai, AWS, and others went all-in on inference at the edge.e

Solid predictions all around – Matthew takes the crown!

06:08 Jonathan – “That’s good; it only took us 6 years to know what the hell we’re talking about!” 

06:23 2025 Stats Review

• We covered 1,308 stories from 15 different, unique sources.
• Amazon accounted for 39% of those stories.
• Ryan’s favorite, Azure, made up 22.9% of the stories (Thanks, Matt…) 
• GCP was 38.1% of our news announcements. 
• The official blogs from cloud providers, including AWS, Azure, and GCP, made up the bulk of the sources for the above stories. 
• This is an interesting change from the first year we recorded, 2019, when AWS accounted for 73% of the announcements. 
• When it comes to host participation, only 6 shows had all four hosts participating. Justin was present for 95%, Ryan for 85%, Matt recorded 78% (not bad with a new baby, honestly), and we had Jonathan for 12 episodes. 
• We only had one guest, and increasing the number of guests is one of our 2026 resolutions, so thanks to Elise for joining us. 
• AI was mentioned 526 times, averaging 12.2x per episode (which seems low to the show note editor), and has definitely been growing each year exponentially. 
• Outages were discussed 19 times (boooo). 
• And we got to talk about our favorite topic, deep-sea cables, 5 times. 
• There were 58.9 hours of runtime over the course of 49 shows, with an average length of 72 minutes.
• The in memorium includes AWS Cloud Search, Glacier, Migration Hub, S3 Object Lambda, Azure Consumption API, dial-up internet, and RC4 encryption, among many others. RIP.  
• The most mentioned non-hyperscaler company was OpenAI, followed closely by Nvidia and Antropic. 
• Lastly, Justin has updated our show LLM Bolt, building a brand new data pipeline for the podcast, which will include show notes, transcripts, etc., all with a new AI-based search. Want to check it out? Join our Slack channel! 

16:28 Ryan – “I’m having a similar experience mostly in my day job… trying to use AI for different workloads and then falling back into more traditional technologies or different ways, and at first I thought it was just like old dog, new tricks, just falling back in the comfort zone. But I find more and more I’m identifying things that, you know, the large language models just are not good at. And I think a lot of stats and the metrics, it feels like it should be able to do that, right? Because it’s conversational and you’re building a corpus of data for the model to query and do all that, but that it really can’t, right? And so, fortunately, we do have machine learning technologies and the ability to do notebooks and stuff. And agentic can absolutely help you make the notebook, but it can’t do the analysis for you, which I find funny.”

To be a good vibe coder, you need to be an experienced programmer, you need to have business experience, and I don’t think the people who are vibe coding right now are getting really good results if they don’t have that kind of background.” 

https://tcp-media.s3.us-west-2.amazonaws.com/2025_year_in_review.html 

25:54 Favorite Announcements

• • Justin:
    • Amazon saying F*** your security to Microsoft was great. 
      • Episode 287: Recorded for the week of Jan 8th, 2025: The Cloud Pod rebrands to The Cloud AI so we can get a 1B valuation.
      • https://www.csoonline.com/article/3625205/amazon-refuses-microsoft-365-deployment-because-of-lax-cybersecurity.html
    • Episode 303 – Someday You Will Find Me, Caught beneath the AI Landslide, in a Champagne Premier Nova in the Sky, from May 18th.  
      • https://aws.amazon.com/blogs/aws/amazon-nova-premier-our-most-capable-model-for-complex-tasks-and-teacher-for-model-distillation/
    • Episode 288: Recorded for the week of Jan 14th, 2025: You might be able to retrain Notebook LM hosts to be less annoyed, but not your cloud pod hosts
      • https://www.theverge.com/2025/1/6/24337530/nvidia-ces-digits-super-computer-ai 
    • Episode 322: Recorded for September 16th, 2025: Did OpenAI and Microsoft break up?… It’s complicated
      • https://www.anthropic.com/news/claude-4 
  • Matt: 
    • Chime is dead: Update on Support for Amazon Chime 
      • episode 294: “Ding: Chime is Dead”** (recorded for the week of February 25th, 2025).
    • GitHub Will Prioritize Migrating to Azure Over Feature Development – The New Stack
      • Episode 317** (“I Got 99 Problems, But a Hallucination Ain’t One”).
      • https://thenewstack.io/github-will-prioritize-migrating-to-azure-over-feature-development/
    • Claude on Azure
      • **Episode 331** is where Claude’s big Azure announcement happened!
      • The episode title says it all: “Claude Gets a $30 Billion Azure Wardrobe and Two New Best Friends” (published November 18, 2025).
  • Ryan:
    • A2A protocol

• Jonathan:

• • DeepSeek is stirring things up
  • AWS Frontier Agents

47:35 2026 Predictions

• Matt
  • A Major GCP Outage will occur
  • A step forward in quantum computing (A quantum leap into 2026)
  • A new MicroHyperscaler will go into the market at the same level as Digital Ocean
• Justin
  • AI Layoff Regret
  • AI Agent Security Breach (Agent that breaches an organization and exfiltrates data)
  • AI-designed web instead of Eyeballs/Humans
• Ryan
  • Multi-Agent Orchestration will blow up in a big way. Major providers of more A2A integrations of workflows between services/clouds
  • Infrastructure as Code will turn into Infrastructure as Intent. 
  • Full Stack Media Creation company with AI? With CMS and Providence tracking and watermarking. Tooling/etc.
• Jonathan
  • Highly Visible company bankruptcy due to rising AI/GPU/Inference Costs.
  • Explosion of Competition against existing SaaS companies
  • An entirely AI-generated Podcast episode from the cloud pod

56:11 Ryan – “Trying to think through emerging threats on technology that I barely understand – because it’s coming out so fast – it’s changing the way we work. You’re already starting to see AI in attacks where groups of people are using AI to put together pretty sophisticated attacks on companies. It’s a lot easier for natural language speakers to generate content for spearfishing; it’s a lot easier for malicious actors to have an AI agent to do a bunch of research on a company real quick, and this is where I think it will be weak.” 

Closing

And that is the week in the cloud! Visit our website, the home of the Cloud Pod, where you can join our newsletter, Slack team, send feedback, or ask questions at theCloudPod.net or tweet at us with the hashtag #theCloudPod

Download audio: https://episodes.castos.com/5e2d2c4b117f29-10227663/2320455/c1e-60w0com436uj0wrx-rk2d6gxvf241-kfvk5f.mp3

Topics covered in this episode:

• port-killer
• How we made Python's packaging library 3x faster
• CodSpeed
• Extras
• Joke

Watch on YouTube

About the show

Sponsored by us! Support our work through:

• Our courses at Talk Python Training
• The Complete pytest Course
• Patreon Supporters

Connect with the hosts

• Michael: @mkennedy@fosstodon.org / @mkennedy.codes (bsky)
• Brian: @brianokken@fosstodon.org / @brianokken.bsky.social
• Show: @pythonbytes@fosstodon.org / @pythonbytes.fm (bsky)

Join us on YouTube at pythonbytes.fm/live to be part of the audience. Usually Monday at 11am PT. Older video versions available there too.

Finally, if you want an artisanal, hand-crafted digest of every week of the show notes in email form? Add your name and email to our friends of the show list, we'll never share it.

Michael #1: port-killer

• A powerful cross-platform port management tool for developers.
• Monitor ports, manage Kubernetes port forwards, integrate Cloudflare Tunnels, and kill processes with one click.
• Features:
  • 🔍 Auto-discovers all listening TCP ports
  • ⚡ One-click process termination (graceful + force kill)
  • 🔄 Auto-refresh with configurable interval
  • 🔎 Search and filter by port number or process name
  • ⭐ Favorites for quick access to important ports
  • 👁️ Watched ports with notifications
  • 📂 Smart categorization (Web Server, Database, Development, System)

Brian #2: How we made Python's packaging library 3x faster

• Henry Schreiner
• Some very cool graphs demonstrating some benchmark data.
• And then details about how various speedups
  • each being 2-37% faster
  • the total adding up to about 3x speedup, or shaving 2/3 of the time.
• These also include nice write-ups about why the speedups were chosen.
• If you are trying to speed up part of your system, this would be good article to check out.

Michael #3: AI’s Impact on dev companies

• On TailwindCSS: via Simon
  • Tailwind is growing faster than ever and is bigger than it has ever been
  • Its revenue is down close to 80%.
  • 75% of the people on our engineering team lost their jobs here yesterday because of the brutal impact AI has had on our business.
  • “We had 6 months left”
  • Listen to the founder: “A Morning Walk”
  • Super insightful video: Tailwind is in DEEP trouble
• On Stack Overflow: See video.
  • SO was founded around 2009, first month had 3,749 questions
  • December, SO had 3,862 questions asked
  • Most of its live it had 200,000 questions per month
  • That is a 53x drop!

Brian #4: CodSpeed

• “CodSpeed integrates into dev and CI workflows to measure performance, detect regressions, and enable actionable optimizations.”
• Noticed it while looking through the GitHub workflows for FastAPI
• Free for small teams and open-source projects
• Easy to integrate with Python by marking tests with @pytest.mark.benchmark
• They’ve releases a GitHub action to incorporate benchmarking in CI workflows

Extras

Brian:

• Part 2 of Lean TDD released this morning, “Lean TDD Practices”, which has 9 mini chapters.

Michael:

• Our Docker build just broke because of the supply chain techniques from last week (that’s a good thing!). Not a real issue, but really did catch an open CVE.
• Long passwords are bad now? ;)

Joke: Check out my app!

Download audio: https://pythonbytes.fm/episodes/download/465/stack-overflow-is-cooked.mp3

Trusted Signing is now Artifact Signing—and it’s officially Generally Available! 

Artifact Signing is a fully managed, end-to-end code signing service that makes it easier than ever for Windows application developers to sign their apps securely and efficiently. 

As Artifact Signing rebrands, customers will see changes over the next weeks. Please refer to our Learn docs for the most updated information. 

Figure 1: Artifact Signing information

What is Artifact Signing? 

Code signing has traditionally been a complex and manual process. Managing certificates, securing keys, and integrating signing into build pipelines can slow teams down and introduce risk. Artifact Signing changes that by offering a fully managed, end-to-end solution that automates certificate management, enforces strong security controls, and integrates seamlessly with your existing developer tools. With zero-touch certificate management, verified identity, role-based access control, and support for multiple trust models, Artifact Signing makes it easier than ever to build and distribute secure Windows applications. Whether you're shipping consumer apps or internal tools, Artifact Signing helps you deliver software that’s secure.  

Security Made Simple 

Zero-Touch Certificate Management  

No more manual certificate handling. The service provides “zero-touch” certificate management, meaning it handles the creation, protection, and even automatic rotation of code signing certificates on your behalf. These certificates are short-lived and auto renewed behind the scenes, giving you tighter control, faster revocation when needed, and eliminating the risks associated with long-lived certs. Your signing reputation isn’t tied to a single certificate. Instead, it’s anchored to your verified identity in Azure, and every signature reflects that verified identity. 

Verified Identity 

Identity validation with Artifact Signing ensures your app’s digital signature displays accurate and verified publisher information. Once validated, your identity details, such as your individual or organization name, are included in the certificate. This means your signed apps will show a verified publisher name, not the dreaded “Unknown Publisher” warning. 

The entire validation process happens in the Azure portal. You simply submit your individual or organization details, and in some cases, upload supporting documents like business registration papers. Most validations are completed within a few business days, and once approved, you’re ready to start signing your apps immediately. 

Figure 2: New organization validation pageFigure 3: Verified ID Process

Secure and Controlled Signing (RBAC) 

Artifact Signing enforces Azure’s Role-Based Access Control (RBAC) to secure signing activities. You can assign specific Azure roles to accounts or CI agents that use your Artifact Signing  resource, ensuring only authorized developers or build pipelines can initiate signing operations. This tight access control helps prevent unauthorized or rogue signatures. 

Full Telemetry and Audit Logs 

Every signing request is tracked. You can see what was signed, when, and by whom in the Azure portal. This logging not only helps with compliance and auditing needs but also enables fast remediation if an issue arises. For example, if you discover a particular signing certificate was used in error or compromised, you can quickly revoke it directly from the portal. The short-lived nature of certificates in Artifact Signing further limits the window of any potential misuse. 

Artifact Signing gives you enterprise-grade security controls out of the box: strong protection of keys, fine-grained access control, and visibility. For developers and companies concerned about supply chain security, this dramatically reduces risk compared to handling signing keys manually. 

Built for Developers  

Artifact Signing was built to slot directly into developers’ existing workflows. You don’t need to overhaul how you build or release software, just plug Artifact Signing into your toolchain: 

Figure 4: Plugging artifact signing into your toolchain

• GitHub Actions & Azure DevOps: The service includes first-class support for modern CI/CD. An official GitHub Action is available for easy integration into your workflow YAML, and Azure DevOps has tasks for pipelines. With these tools, every Windows app build can automatically sign binaries or installers—no manual steps required. Since signing credentials are managed in Azure, you avoid storing secrets in your repository. 

Figure 5: Example of the build output of a wpf application

• Visual Studio & MSBuild: Use the Artifact Signing client with SignTool to integrate signing into publish profiles or post-build steps. Once the Artifact Signing client is installed, Visual Studio or MSBuild can invoke SignTool as usual, with signatures routed through the Artifact Signing service. 

• SignTool / CLI: Developers using scripts or custom build systems can continue using the familiar signtool.exe command. After a one-time setup, your existing SignTool commands will sign via the cloud service. The actual file signing on your build machine uses a digest signing approach: SignTool computes a hash of your file and sends that to the Artifact Signing service, which returns a signature. The file itself isn’t uploaded, preserving confidentiality and speed. This way, integrating Artifact Signing can be as simple as adding a couple of lines to your build script to point SignTool at Azure. 

• PowerShell & SDK: For advanced automation or custom scenarios, Artifact Signing supports PowerShell modules and an SDK. These tools allow you to script signing operations, bulk-sign files, or integrate signing into specialized build systems. 

The Right Trust for the Right Audience 

Artifact Signing has support for multiple trust models to suit different distribution scenarios. You can choose between Public Trust and Private Trust for your code signing, depending on your app’s audience: 

• Public Trust: This is the standard model for software intended to go to consumers. When you use Public Trust signing, the certificates come from a Microsoft CA that’s part of the Microsoft Trusted Root Program. Apps signed under Public Trust are recognized by Windows as coming from a known publisher, enabling a smooth installation experience when security features such as Smart App Control and SmartScreen are enabled.  

• Private Trust: This model is for internal or enterprise apps. These certificates aren’t publicly trusted but are instead meant to work with Windows Defender Application Control (App Control for Business) policies. This is ideal for line-of-business applications, internal tools, or scenarios where you want to tightly control who trusts the app. Artifact Signing ’s Private Trust model is the modern, expanded evolution of Microsoft’s older Device Guard Signing Service (DGSS) -- delivering the same ability to sign internal apps but with ease of access and expanded capabilities.  

• Test Signing: Useful for development and testing. These certificates mimic real signatures but aren’t publicly trusted, allowing you to validate your signing setup in non-production environments before releasing your app. 

Note on Expanded Scenario Support: Artifact Signing supports additional certificate profiles, including those for VBS enclaves and Private Trust CI Policies. In addition, there is a new preview feature for signing container images using the Notary v2 standard from the CNCF Notary project. This enables developers to sign Docker/OCI container images stored in Azure Container Registry using tools like the notation CLI, backed by Artifact Signing.

Having all trust models in one service means you can manage all your signing needs in one place. Whether your code is destined for the world or just your organization, Artifact Signing makes it easy to ensure it is signed with an appropriate level of trust. 

Misuse and Abuse Management 

Artifact Signing is engineered with robust safeguards to counter certificate misuse and abuse. The signing platform employs active threat intelligence monitoring to continuously detect suspicious signing activity in real time. The service also emphasizes prevention: certificates are short-lived (renewed daily and valid for only 72 hours), which means any certificate used maliciously can be swiftly revoked without impacting software signed outside its brief lifetime. When misuse is confirmed, Artifact Signing quickly revokes the certificate and suspends the subscriber’s account, removing trust from the malicious code’s signature and stopping further abuse. These measures adhere to strict industry standards for responsible certificate governance. By combining real-time threat detection, built-in preventive controls, and rapid response policies, Artifact Signing gives Windows app developers confidence that any attempt to abuse the platform will be quickly identified and contained, helping protect the broader software ecosystem from emerging threats. 

Availability and What’s Next 

Check out the upcoming “What’s New” section in the Artifact Signing Learn Docs for updates on supported file types, new region availability, and more. Microsoft will continue evolving the service to meet developer needs. 

Conclusion: Enhancing Trust and Security for All Windows Apps 

Artifact Signing empowers Windows developers to sign their applications with ease and confidence. It integrates effortlessly into your development tools, automates the heavy lifting of certificate management, and ensures every app carries a verified digital signature backed by Microsoft’s Certificate Authorities. 

For users, it means peace of mind. For developers and organizations, it means fewer headaches, stronger protection against supply chain threats, and complete control over who signs what and when. 

Now that Artifact Signing is generally available, it’s a must-have for building trustworthy Windows software. It reflects Microsoft’s commitment to a secure, inclusive ecosystem and brings modern security features like Smart App Control and App Control for Business within reach, simply by signing your code. Whether you're shipping consumer apps or internal tools, Artifact Signing helps you deliver software that’s both easy to install and tough to compromise. 

Thank you to everyone who participated in Advent of Code 2025 in Kotlin! Once again, it was a joy to see so many of you sharing solutions, cheering each other on in Slack, and keeping the holiday coding spirit alive.

This year, we kicked things off with five days of livestreams on December 1–5, with Sebastian Aigner and fantastic guests solving puzzles live, discussing strategies, and showing off idiomatic Kotlin approaches.

If you missed the streams or want to revisit a clever trick, you can still catch up with the recordings:

Community

The Advent of Code puzzles remain available all year round, and so does the community energy. The #advent-of-code channel in the Kotlinlang Slack was once again full of:

• Creative Kotlin snippets
• Non-spoiler hints
• Lots of encouragement

Whether you took part in discussions every day or just lurked and learned, thank you for making the channel such a friendly, collaborative space.

Join the AoC channel in Slack

Winners

As in previous years, we invited you to solve Advent of Code puzzles in Kotlin, join our Kotlin leaderboards, and share your solutions on GitHub with the aoc-2025-in-kotlin topic for a chance to win special Kotlin prizes.

This year, we’re celebrating nine winners across three categories: Fastest to Solve, Random, and Community.

Fastest to Solve

These three contestants topped the combined Kotlin leaderboards, consistently solving the puzzles at impressive speed:

• Kroppeb
• AndrejStratmann
• 770grappenmaker
  

Congratulations on those lightning-fast stars and Kotlin-powered solutions!

Random winners

To give everyone a chance regardless of speed, we also randomly selected three prize winners from all the participants:

• lbcp
• FelixDombek
• bjdupuis

If you see your name here, keep an eye on your inbox. We’ll be in touch about your prize soon.

Community winners

Finally, Advent of Code in Kotlin wouldn’t be the same without the people who answer questions, share insights and alternative solutions, help newcomers get started, and keep the #advent-of-code channel buzzing throughout December.

This year, we’d like to give a special shout-out to these three community stars:

• jakubgwozdz
• bj0
• ephemient
  

Thank you for your team spirit in the Kotlin community!

Keep learning and solving

Advent of Code might be over for this year, but the puzzles and the learning opportunities remain. If you’d like to keep sharpening your skills and prepare for future events, here are some resources:

• Advent of Code 2025 in Kotlin YouTube playlist.
• Solutions to Advent of Code puzzles from previous years in idiomatic Kotlin.
• A tutorial on how to explore and solve programming puzzles using Kotlin Notebooks.

We want to say a huge thank-you, as always, to Eric Wastl and the Advent of Code team for creating such a beloved set of puzzles year after year.

Let’s keep exploring Kotlin, solving algorithmic challenges, and supporting each other. We hope to see you for the next Advent of Code in Kotlin! 🎄

Today, a quick example of introducing determinism when an AI struggles.

I have a simple instruction for it:

Run `date` to identify the current week. Check the calendar for the week to identify workdays, travel, and meetings

I tell it to run date because it’s terrible at knowing what day it is. In its CLAUDE.md, I gave it a command for checking the calendar.

Today is Monday. What does it say?

Calendar summary:

• Monday Jan 12 (TODAY): No meetings showing in your calendar – open day!
• Tuesday Jan 13: Heavy meetings (10am-5pm solid)
• Wednesday Jan 14: Moderate meetings + evening at Avdi’s
• Thursday Jan 15: Moderate meetings
• Friday Jan 16: Some focus work time (2:15-4pm blocked), meetings, evening events

Ha ha I wish!

It’s doing it wrong! After yelling at my computer, I set about fixing this for the future.

Step 1: Ask why it’s wrong.

It can tell me what it did and notice what went wrong. In this case, I have a pretty good idea where to point it.

Me: what command did you use for gcalcli? because it is not working
Claude: Ah, I see the problem! I used:

gcalcli --calendar jessitron@honeycomb.io "Monday Jan 13" "Sunday Jan 19" 

But Monday is Jan 12, not Jan 13! So I missed today entirely. Let me fix that…

Yet again, the LLM proves bad at math. Especially dates.

Step 2: Introduce determinism

I want to fix this problem permanently. Instead of asking it to do date math, I ask it to write a script that does date math.

Me: ok, let’s make a script. Make a script for week-agenda, and put make it
calculate the start and end date as monday through friday. Then it calls
gcalcli

Claude: Good idea!

It writes a script. I try the script. It left out Friday this time. I tell it so, and it fixes the script.

Next I tell it to change its own instructions. No more trying, only succeeding, from now on.

Me: Now update your skill to point to the script
Claude: Good idea. Let me update the skill to use the script.

Finally, my favorite part: /clear. Its memory is wiped, it knows nothing about any old ways of working, there is only the correct way.

A bit of deterministic code goes a long way.

The weaknesses of AI can be obviated by its strengths. It’s bad at date math, but it’s good at writing code. A little direction from me, and it gets better at its job all the time. This is working with AI.