In the tech world, there is a constant flow of changes and keeping up with them means the choice for tools and technologies which are the most appropriate to invest your time in. In 2025 the best programming language or technology stack to learn really depends on your personal aims, hobbies, and apps you are going to create.
The interest in Java is dropping. February 2025 TIOBE programming community index. C++, which has long been the cornerstone of system programming and performance-critical applications, has officially overtaken Java to take second place in the TIOBE programming language popularity index. A new report from the Java vendor Azul claims that 88% of companies are considering moving off of Oracle Java to another alternative as a result of rising costs and restrictive policies from Oracle, among other issues.
The growing trend in the world of software development: speed matters. C++, Go, and Rust are gaining popularity because the need for computing power increases faster than speed of CPUs is increasing, sothere is a growing interest to the fast programming languages. While C++ is establishing itself, other fast languages are making significant strides. Go continues its top 10 ranking, while Rust has reached an all-time high.
Python still holds its place at the top of the programming world. Since the number of trained experts in the software industry is not enough to cover the growing need, professionals from many other fields are taking over programming skills with the help of Python. This ensures that Python maintains its position even as speed continues to be emphasized in programming language choices. Programs written with Python are often notoriously slow and inefficient. Python 3.14, due out later this year, is set to receive a new type of interpreter that can boost performance by up to 30% with no changes to existing code. Write Python like it’s 2025 and check Python Libraries That Will Make You Feel Like a Data Wizard.
There are also innovative alternatives to the popular languages are gaining steam—and one of them could be the perfect fit for your next project. Top programming languages to learn in 2025: Python, JavaScript, Rust, and more – maybe also Go. Check out also those 11 cutting-edge programming languages to learn now or decide it is better for you to not going to learn a new programming language this year.
Microsoft is actively pushing Visual Studio Code extensions for many uses and even replacing existing separate tools. GitHub Copilot is advertised as your AI pair programmer tool in Visual Studio Code. Check the Best VS Code Extensions to Boost Your Productivity.
Best Backend Frameworks for 2025: A Developer’s Guide to Making the Right Choice The stakes for choosing the right backend framework have never been higher. With the explosion of AI-powered applications, real-time processing requirements, and microservices architectures, your framework choice can make or break your project’s success.
Artificial intelligence (AI) is accelerating at an astonishing pace, quickly moving from emerging technologies to impacting coding a lot AI tools have come heavily to the coding. Coders use AI to help their coding in many ways. You can write code quickly. How to refactor code with GitHub Copilot. How To Build Web Components Using ChatGPT. There are also warnings that Using GitHub Copilot is one sure-fire way to never actually learn how to do coding.
The web has come a long way from static HTML pages to dynamic and highly interactive applications. When traditional JavaScript-based web apps struggle with performance-intensive tasks, WebAssembly (WASM) promises to enable near-native performance on the web. Read Why WebAssembly (WASM) is the Future of High-Performance Web Apps.
JavaScript in 2025 will see advancements in serverless architectures, integration with WebAssembly, adoption of microfrontends, and more. JavaScript is also a fighting field. Deno filed a petition with the United States Patent and Trademark Office to cancel Oracle’s trademark in November 2024. Oracle will not voluntarily release its trademark on the word “JavaScript”. Building Modern React Apps in 2025 – A Guide to Cutting-Edge Tools and Tech Stacks
The open source, cross-platform JavaScript runtime environment Node.js will soon support TypeScript by default, without extra configuration. Node 23 will be able to run TypeScript files without any extra configuration. Express is an extremely commonly used web server application framework in Node.js.
Open Source in 2025: Strap In, Disruption Straight Ahead article takes a look for new tensions to arise in the New Year over licensing, the open source AI definition, security and compliance, and how to pay volunteer maintainers. For good news check out Top Open Source Projects to Watch in 2025 and 13 top open-source tools you must use for your next big project in 2025.
The Mobile Development Tech Stack for 2025 selection is important because the right tech stack can make or break your mobile app. The mobile development tech stack for 2025 is rich with opportunities.
Must-Know 2025 Developer’s Roadmap and Key Programming Trends article says that in the world of coding trends, one thing is clear: classic languages like Java, Python, and JavaScript are still important, but they’re being joined by new favorites such as Go and Rust. And when you ask “Is JavaScript or Python 2025?” the answer is rarely simple – and could be that you need both.
Here are some points:
Python’s Growth in Data Work and AI: Python continues to lead because of its easy-to-read style and the huge number of libraries available for tasks from data work to artificial intelligence. Tools like TensorFlow and PyTorch make it a must-have.
JavaScript and Its Ongoing Role in Building Website: JavaScript (and by extension, TypeScript) is the basic building block. JavaScript is still essential for web work, running both the parts you see on a site and the behind-the-scenes work, but many coders are now preferring TypeScript for business projects. Try building a small web app using React.
The Rise of Go and Rust: For those looking at future coding languages 2025, Go and Rust are getting a lot of attention.
Java, C++, and C#: The Reliable Favorites: Even in 2025, there’s no ignoring that languages like Java, C++, and C# are still important. Java continues to be a top choice for large business applications and Android app development, while C++ is key in systems work and game development.
There are several shifts that every aspiring coder should keep in mind:
Adding Artificial Intelligence to Coding: The future of coding is closely linked with AI
Building for the Cloud: With cloud computing becoming common, languages that handle many tasks at once and run fast (like Go and Rust) are more important than ever.
The Need for Full-Stack Skills: Coders today are expected to handle both the front part of websites and the back-end work. JavaScript, along with tools like Node.js and modern front-end libraries, is key.
Focus on Safety and Speed: With online security becoming a big issue, languages that help avoid mistakes are getting more attention. Rust’s features that prevent memory errors and Go’s straightforward style are good examples.
Keep Learning and Stay Flexible: One thing that never changes in tech is change itself. What is popular in 2024 might be different in 2025.
Here’s a simple table that sums up some facts in plain language:
| Language | 2025 Trend | Main Advantage | Resource Link |
|---|---|---|---|
| Python | Leads in data work and AI | Easy to read, lots of tools | GeeksforGeeks |
| JavaScript | Essential for building websites | Works everywhere on the web | Snappify |
| TypeScript | Becoming popular in large projects | Helps catch errors early | Fullstack Academy |
| Go | Growing quickly in cloud computing | Fast and handles many tasks at once | Nucamp |
| Rust | New favorite for safe, low-level coding | Prevents common memory mistakes | The Ceres Group |
| Java | Still important for big business and Android work | Runs on many types of systems | Wikipedia |
Best Dev Stacks to Learn in 2025lists the top development stacks for 2025 to be:
1. MERN Stack (MongoDB, Express.js, React, Node.js)
2. MEVN Stack (MongoDB, Express.js, Vue.js, Node.js)
3. JAMstack (JavaScript, APIs, Markup)
4. T3 Stack (Next.js, TypeScript, tRPC, Tailwind CSS, Prisma)
5. Flutter Stack (Flutter, Firebase)
6. PERN Stack (PostgreSQL, Express.js, React, Node.js)
7. Django Stack (Django, PostgreSQL, React/Angular)
8. DevOps Stack (Docker, Kubernetes, Jenkins, Terraform)
9. AI/ML Stack (Python, TensorFlow, PyTorch, FastAPI)
10. Blockchain Development Stack (Solidity, Ethereum, Hardhat)
11. Spring Boot + React Stack
10 hot programming trends — and 10 going cold
Hot: Repatriation
Not: Cloud bills
Hot: AI partners
Not: Human pair programming
Hot: Rust
Not: C/C++
Hot: Wasm
Not: Interpreters
Hot: CPUs
Not: GPUs
Hot: Zero-knowledge proofs
Not: Digital signatures
Hot: Trustworthy ledgers
Not: Turing-complete ledgers
Hot: GraphQL
Not: REST
Hot: Static site generators
Not: Single-page apps
Hot: Database configuration
Not: Software programming
What’s trending in Software-driven Automation (SDA) in 2025? Here are some predictions:
1. Virtual Safe Control – A new and novel concept introduced by CODESYS and SILista, making it possible to implement Functional Safety controller reaching SIL2 or even SIL3 level, using generic hardware with help of software virtualisation. This will significantly decrease cost of hardware and speed up development cycle.
2. Open platforms – This trend started already last year, and now we’re seeing more and more automation vendors coming this way. #ctrlXOS opened the game, and there are other vendors like Phoenix coming the same way with their PLCnext Virtualised.
3. Model-based Design (MBD) – An old concept but not yet fully utilised in development. Maybe because lack of well integrated toolchains in the past. But now we’re seeing more and more industrial players adopting the methodology in their product development.
4. AI, of course, but how? Naturally AI can assist in efficient software development and testing. Also some algorithm optimisation and condition monitoring with AI and ML has been seen.
663 Comments
Tomi Engdahl says:
https://www.xda-developers.com/self-hosted-tool-to-use-my-home-network-while-im-traveling/
Tomi Engdahl says:
https://dev.to/therealmrmumba/my-fav-open-source-github-tools-2025-as-a-developer-2o6b
Tomi Engdahl says:
https://towardsdatascience.com/agents-apis-and-the-next-layer-of-the-internet/
Tomi Engdahl says:
https://github.blog/open-source/git/highlights-from-git-2-50/
Tomi Engdahl says:
https://infosecwriteups.com/how-i-automated-my-entire-infrastructure-with-one-tool-and-saved-20-hours-a-week-59fd6020b38f
Tomi Engdahl says:
https://www.kdnuggets.com/10-github-repositories-to-master-web-development-in-2025
Tomi Engdahl says:
https://github.com/telit/IoT-AppZone-SampleApps/tree/master/ME310G1-ME910G1-ML865G1
Tomi Engdahl says:
https://analyticsindiamag.com/ai-features/the-story-of-a-prisoner-who-became-a-software-engineer/
The Story of a Prisoner Who Became a Software Engineer
“I’m very grateful that LLMs are something that I did not have available to me for a large portion of my time learning
Tomi Engdahl says:
Beyond Code Generation: Continuously Evolve Text with LLMs
Long-running content evolution and an introduction to result analysis
https://towardsdatascience.com/beyond-code-generation-continuously-evolve-text-with-llms/
Tomi Engdahl says:
Want to supercharge your vibe coding skills? Here are the best AI models developers can use to generate secure code
Claude 3.7 Sonnet is the best performer for vibe coding, while others produce very mixed results
https://www.itpro.com/software/development/vibe-coding-best-ai-models-secure-code-generation
Vibe coding has become the latest big trend in software development, with devs ramping up the use of AI tools to automate code generation.
Tomi Engdahl says:
Can vibe coding produce production-grade software?
https://www.thoughtworks.com/insights/blog/generative-ai/can-vibe-coding-produce-production-grade-software
We’ve been discussing the concept of ‘vibe coding’ a lot at Thoughtworks recently. But can it actually be used to write software that we put out into the world? Prem Chandrasekaran did three experiments to see what would work and what wouldn’t.
The idea of letting an AI write production-grade code can stir both fascination and doubt. Some see the promise of near-instant productivity — code at the click of a button — while others worry about unleashing legions of barely readable and unmaintainable scripts into our codebases. As practitioners who have spent countless hours refining standards of “good” code, we approached this debate with a mix of curiosity and caution.
What do we mean by “production-grade” software?
Before diving into the experiments, it’s worth acknowledging that production-grade is not a universally defined term. If there were a single, widely accepted definition, many real-world codebases might struggle to meet it.
In practice, the term reflects a combination of engineering judgment, organizational context and lived experience. For the purposes of this article, we don’t use production-grade to imply perfection, but rather to describe software that’s robust, maintainable and ready to be deployed and evolved responsibly in real-world conditions.
To keep things grounded, we roughly eyeballed a mix of qualitative indicators and quantitative signals that experienced developers often associate with production-worthy code. These aren’t hard rules, but directional heuristics — enough to assess whether the software feels trustworthy, evolvable and operationally sound.
Tomi Engdahl says:
A glimpse into the future of software development
AI tools are no longer just glorified auto-completers or snippet generators. They are increasingly capable of building coherent, end-to-end solutions — APIs, databases, tests, everything. Yes, they still make mistakes. Yes, they still require human supervision. But the bar has been raised.
What once took a team days can now be scaffolded by a single developer in hours — with an AI partner moving at breathtaking speed. This no longer feels like a novelty — it’s a paradigm shift in how software is built.
So, can AI-assisted coding produce production-grade software? Not consistently — not yet.
But the gap is closing. With the right architectural intent, oversight and feedback loops, AI is inching closer to becoming a reliable teammate.
At the same time, it’s important to acknowledge: today’s AI models do not inherently optimize for self-verifiable code — code that asserts its correctness through automated tests, assertions, or contracts.
https://www.thoughtworks.com/insights/blog/generative-ai/can-vibe-coding-produce-production-grade-software
Tomi Engdahl says:
Claude Code saved us 97% of the work — then failed utterly
Experimenting with Claude Code in CodeConcise
https://www.thoughtworks.com/insights/blog/generative-ai/claude-code-codeconcise-experiment
Tomi Engdahl says:
https://www.geeky-gadgets.com/ai-rewriting-its-own-code/
Tomi Engdahl says:
Want to supercharge your vibe coding skills? Here are the best AI models developers can use to generate secure code
Claude 3.7 Sonnet is the best performer for vibe coding, while others produce very mixed results
https://www.itpro.com/software/development/vibe-coding-best-ai-models-secure-code-generation
Tomi Engdahl says:
https://www.kdnuggets.com/go-vs-python-for-modern-data-workflows-need-help-deciding
Tomi Engdahl says:
https://www.infoq.com/news/2025/06/void-ide-beta-release/
Tomi Engdahl says:
https://www.businessinsider.com/google-engineer-advice-students-internships-first-second-year-programs-2025-6
Tomi Engdahl says:
https://thenewstack.io/rust-eats-pythons-javas-lunch-in-data-engineering/
Tomi Engdahl says:
“You can’t do worse than learn to code.”
Risk Expert Says “Learn to Code” Is Now Worse Advice Than “Get a Face Tattoo”
“You can’t do worse than learn to code.”
https://futurism.com/risk-expert-learn-to-code-face-tattoo?fbclid=IwZXh0bgNhZW0CMTEAAR7WG0gGru9YW4aivONlayiO44gxzn1I1jeyr9HZ675trMYvMzkmI9SuTkflow_aem_tPOwfrMdwTD6uujnYnB9IQ
During a recent exchange on “Real Time with Bill Maher,” risk analyst Ian Bremmer argued that artificial intelligence has basically eviscerated the “learn to code” cottage industry.
When discussing the way AI has swept the world and taken many white collar jobs with it, Bremmer referenced how rapidly the technology has overtaken the traditional career trajectory for programmers — so much so that people who used to have cushy software developer jobs are now selling their plasma to make ends meet.
“Just five years ago, the smartest advice that we had for the kids was ‘learn how to code,’” Bremmer recollected. “That is literally worse advice now than ‘get a face tattoo.’ You can’t do worse than learn to code.”
Indeed, in its latest labor market report, the New York Federal Reserve found that recent college graduates who majored in computer science or computer engineering have higher rates of unemployment than those who studied journalism, political science, and even English.
Overall, computer science majors had the seventh-highest rate of unemployment at 6.1 percent, and computer engineering majors had the third-highest at 7.5 percent, per the New York Fed. Compared to the overall recent graduate unemployment rate of 5.8 percent, that’s an embarrassing reversal indeed — especially for a field that was once considered a safe and lucrative bet for prospective students.
During the lively discussion, Bremmer’s fellow guest, historian and author Rutger Bregman invoked a popular silver lining talking point about AI hoovering up white collar jobs: that ultimately, capitalists will “come up with new bullsh*t jobs” for people who lost work to the technology.
In the wake of so many tech layoffs, the new adage is now “learn AI” — but as tech founder Joe Procopio wrote in an Inc magazine column earlier this year, that advice probably about as effective long-term as “learn to code.”
“We’ve already inadvertently created a class of ‘AI talent’ that knows how to code with GitHub Copilot,” the serial entrepreneur and advisor wrote. “This is not going to create better code for better apps for better business outcomes.”
Instead, as Procopio aptly noted, “it’s doing a great job creating AI slop but with code — just like ‘Learn To Code’ created a workforce of terrible coders, who are now easy targets to be replaced by AI slop coders.”
“Damn this vicious cycle,” he concluded.
Tomi Engdahl says:
“Learn to Code” Backfires Spectacularly as Comp-Sci Majors Suddenly Have Sky-High Unemployment
“Every kid with a laptop thinks they’re the next Zuckerberg.”
https://futurism.com/computer-science-majors-high-unemployment-rate
As Newsweek reports, recent college graduates who majored in computer science are facing high unemployment rates alongside the increasing probability of being laid off or replaced by artificial intelligence if and when they do get hired.
Tomi Engdahl says:
Few years later: “why can’t we find enough coders and IT workers?” (Story already seen in 2000s)
Tomi Engdahl says:
Fabrizio Canone there is still a shortage of good engineers
Fabrizio Canone in the 90s they were saying object oriented programming would put most of us out of work. Yep
Fabrizio Canone ok boomer. They didn’t have advanced computer learning language models, and other AI back in the 2000s. Are you talking pre-911? Wtf are you on about?
But there were GUIs to build softwares without even write words. Only drag&drop. And they were advancing fast. Also scripted code generators/converters
Idk, low hanging fruit like entry level coding is probably out now as a side hustle, but serious deep understanding of any progranming language is about to be extremely important as people make more code with AI and need to troubleshoot the bugs in code that no human being has ever laid eyes on. Thats a big task.
Matthew McCoul last year A.I. couldn’t code. By end of year it was in the top 99% of all coders. You think a year from now it won’t outclass everyone?
Major Jackson considering the trends in image generation and LLMs, yes. Yes I think by June 19th of 2026 code AI will be improved, and struggling to improve further because of feedback loops created by the AI being trained on AI-generated code. I think coders will be using it extensively, and still need to do manual coding to make it all work.
I’m marking my calendar so i can find this and reply when the day comes. If AI has made programmers obselete I will mail you a crisp $100 bill.
Robert Jarlaczyk exactly. I use CoPilot as a reference. You have to know what to ask it in the first place. I’ve not seen it work well on anything but really small, focused tasks. Right now it just is a productivity enhancement because it saves skilled engineers time looking things up.
Major Jackson i mean, yes. I think i said as much right in my post. Low level coding will be out. A few experts will be in demand.
Matthew McCoul machine generated code is a mess…..
All code should go through code reviews to address efficiency and vulnerabilities
I bet that was also said when No-code or Low-code was invented
Saulo Gomes Yes. They said this in the 90s about object oriented programming
I remember a few years ago they told people that driving trucks would be replaced very soon.
It’s still very much yet to be seen how much AI will affect the market for programmers. So far, it has only eliminated the lowest of the low entry level developer jobs while jobs in Machine Learning were created at the same time. If you’re already studying Computer Science, I’d encourage you to continue regardless. Even in the worst case scenario, there are a lot of other jobs in tech you can get with a computer science degree.
I think there’s also something to be said that no code and low code editors have existed since the late 1990s and job growth for web developers continued to explode and even bounced back after the dotcom bubble burst.
Jonathan Charles Mitchell not even that – it makes actual programmers more valuable fixing the dog water that novices create with AI
Jonathan Charles Mitchell It’s pretty predictable really. Things are going to change drastically in the next 5 years. If your job is in front of a computer screen doing entry level stuff (90% of people) You’re in trouble.
That’s very skewed, coding for generative AI development was a short term push as any who have been around were saying. Coding knowledge is still needed to create functional systems. AI is helpful, but still has many downfalls. It’s like saying you don’t need mechanics anymore because cars are made differently now than in the past. Still need them, skills change.
2 or 3 years ago it was like reading was back in the middle ages, if you knew how to code you were elite, ahead of your time..
Software development is still the only field that you can work from home, 6-7h a day not the crazy 9 to 10h a day most fields do. Also it is the only field with 0 riscs and low stres.
I’d rather be homeless than doing anything else, especially manual labor.
Meanwhile about 5% of AI code doesn’t even compile while 30% of it doesn’t work. Not to mention anything complex and AI doesn’t even attempt it.
I was waiting for an AI to do the coding for me as game dev is too time consuming and I am no longer interested in spending such a time but AI is next to useless other than coming up with basic coding, completely useless in figuring out bugs.
It is just a faster stack exchange .
The thing i wonder is ai leans by reading other people’s code as new tech comes out if there are no coders then where is it going to get code to learn from
Was this a manager who thinks he’ll be able type, make me a complete e-commerce system without any bugs, without any competing requirements, without having to sit in hours of meetings talking about pixel density, into ChatGPT and it will work first time without any issues. Lol. Ok.
as technology advanced many majors were created but as we have AI now that can do so many things some majors will be marged together such as Computer science with electronics engineers etc but also new majors will pop up such as AI engineers they will be very heavy on mathematics. these are just few examples i can come up with. Coding was only needed to write actual algorithm for computer now we don’t need coding we can simply write the algo and feed it to computer.
At least understand the code. Blindly depending on an auto regressive model for coding will be the last thing you should do
This is a very misleading article. The comparison between coding and face tattoos is ridiculous. Once you can code, you can code. Just keep learning new skills. It’s the same ( ok it’s slightly different) to the electronics industry, the computer industry, IT, video, sound, etc. etc) I have always read and updated my areas of expertise ferociously. After I had been in a particular technology industry for some 10 years, having gone through learning the hardware, the software, installation, operation, etc, I was talking to an old hand who was stuck with the old hardware assembly, and I said, ” why don’t you learn how to wire it up and configure the software, you would earn a lot more money and it won’t end up killing you ( old gear was very heavy) I will show you how to do it, it’s not that difficult ” his reply was. ” No thank you, I know what i know and I do what I do ” I give up.
Tomi Engdahl says:
Our lives are too short and there is too much to know to be able to remember that “coding” (or “programming,” as I remember we used to call it) was born for the sole purpose of translating human language and logic into ones and zeros.
At first we did it by simply punching cards, then by algebraic operators, and finally by increasingly complex commands; so, given this very obvious trend, it would be rather bizarre – if not downright stupid – not to imagine that one day human language and programming language would converge into a single entity.
The only ones who, at most, could complain about it are the children of the 2000s, innocently imbeciled as they are by a popular narrative that has painted programmers as the heroes of the new world, as caffeine-fueled superheroes projected into the future and not, as they are instead realizing they are, sad misfits not so different than the gray characters in dystopian films like “Brazil” or “1984″.
Comment from https://www.facebook.com/share/1HYhW1saJL/
Tomi Engdahl says:
Windsurf with Claude 3.7 thinking
Tomi Engdahl says:
I’ve been using openrouter.ai API with the AI Engine plugin in WordPress. OpenRouter gives you access to lots of different AI APIs with a single key. Claude Opus 4.0 is pretty good for coding, but you have to be precise about what you want it to do, and what to leave out, because it can add things you didn’t ask for at times even with a low temperature setting. I probly spent about $12 on opus 4.0 to get a successful php script for my site to pass LCP on mobile for Core Web Vitals. The issue was finding out the Jannah theme uses the default wordpress content output for the feature image, then developing a script that worked with that plus being able to modify the featured image HTML when it’s served by FastPixel CDN. LCP on mobile was the only issue on my site, a known issue with Jannah theme, everything else was passing CWV with flying colors. Each API call for Claude Opus 4.0 ranged from 12 cents to 23 cents depending on how complicated it was, or if I pasted in pre existing script for it to modify. ChatGPT on the other hand can cost less than a penny or a few pennies per API call with openrouter
https://www.facebook.com/share/p/19AP5PL6Kn/
Hi.
I’m looking for recommendations of AI-based coding services.
Thanks.
The company pays for a Cursor license for me. It’s nice as a copilot, until it decides to go off the rails (as all AI products do) on occasion. Treating it as a semi-smart autocomplete it’s made me more productive.
First tip, call it vibe coding.
Github copilot maybe?
Don’t even think about it. Complete garbage.
Jan Heisterkamp If you don’t expect too much from it it’s okay to play around with
Get the Cline plugin for VSCode – plug in your API key, pick an expensive model for plan mode and an inexpensive model for act mode. Tell plan mode what you want and which files to read and instruct it to develop a comprehensive plan including code snippets, psuedocode, and a technical analysis. I like to use Gemini pro or o3 for this. Then, flip it to act mode and let a cheaper model carry out the plan. I like to use Gemini flash for this. Other models may have too small a context window for big projects. When the context window fills up, flip back to plan mode and ask it to summarize the current progress in detail and create a new task with that plan as the context, and then your context window disappears and you resume with a new plan based on the most recent summary. Saves token costs that way, I usually reset at around 500k tokens. You can also flip back to plan mode when the act mode agent gets stuck or starts looping or getting lost in diff mismatches. o3 does great plans but you can only use it for your first plan of the run because of the limited context window. Claude is good too but it’s very expensive compared to the alternatives. Mistral and Devstral only seem to be good for code completion in my experience but I’m a newbie. Same with Qwen. Deepseek is good but slow and VERY verbose in reasoning – cheap but uses up a lot of tokens and sometimes loses the big picture, Gemini pays attention to the whole context better.
I wrote a bit about it, how to not spend crazy money with API costs etc https://wuu73.org/blog/guide.html you can even spend $0 using llm7 or pollinations api. Unlimited GPT 4.1 use smarter models to plan then 4.1 in cline to make the edits / files
Put all your critical data on Amazon cloud and check craigslist gigs for any Ole cheap self proclaimed deepseek ai “programmer”…..what could possibly go wrong?
Tomi Engdahl says:
https://openrouter.ai/?fbclid=IwY2xjawLHXVxleHRuA2FlbQIxMQABHqVSsj_moVv2YoakOtNWeIQBBE1M5B5jxSXSQKKzGwSeadvtDRpobaK3_oi5_aem_uQoun_0BvgW_3LEYNa3oCw
Tomi Engdahl says:
https://thatonegamedev.com/cpp/writing-generic-code-in-c/
Tomi Engdahl says:
https://dev.to/shayy/the-7-tools-i-use-every-day-to-build-my-5000-user-saas-12f3
Tomi Engdahl says:
Rethinking Low Code and RAD for the AI Era: Why Triple-View Development Is the Future
https://blogs.embarcadero.com/rethinking-low-code-and-rad-for-the-ai-era-why-triple-view-development-is-the-future/
Rapid Application Development (RAD) and Low Code tools have long aimed to simplify and accelerate software creation. Over the years they became excellent at keeping a visual “drag and drop” configuration interface tightly integrated with the traditional code view and tooling of IDEs.
The proliferation of JavaScript with multiple frameworks and less strict standards made this more challenging. It was easy to generate code with a visual interface but keeping the visual interface and the code views aligned was not so easy. Even in the most modern and successful Low Codes this problem persists, especially as solutions frequently require add-ons that are not visually supported or provided by 3rd parties. Keeping things together requires tight control of visual (and non-visual) frameworks and where they are less proprietary, it gets even more difficult.
Classic RAD and Low Codes Provide a Dual View: Code and Visual Development
Modern Low Code solutions solve this problem a lot better and the need to go outside of the core environment is lower.
However, AI creates a new development paradigm. A prompt-based interface can now generate a lot of code. A new developer “view” into the apps.
The Inspiration from Lovable
Emerging platforms like Lovable (www.lovable.dev) show what’s possible when speed and AI come together. With natural language prompts, users can scaffold working applications in minutes. It’s intuitive and fast—great for getting started or experimenting.
The AI doesn’t remain connected to the code, nor is there an easy path to transition between AI-driven creation and deeper customization. Once the initial magic is done, you’re largely on your own.
The Future: A Triple-View IDE
The future of development tooling must support three fully integrated modes of interaction:
AI Prompt View Start with natural language. Describe what you want—a login page, a dashboard, a workflow—and let AI do the heavy lifting.
Visual Editor View Drag-and-drop still has value. It’s fast for layout, UX design, and understanding application flow.
Code View Developers will always need full control. Whether it’s to optimize performance, customize behavior, or troubleshoot issues, the ability to dive into clean, modular code is essential.
The critical breakthrough is live synchronization across all three. Change something in the code? The visual view and AI prompt history update accordingly. Edit the layout visually? The underlying code refactors cleanly. Adjust the AI prompt? The changes cascade to both visual and code views.
What This Means for RAD and Low Code Tooling
To stay relevant, modern RAD tools must:
Embrace AI as a co-developer, not just a jumpstart.
Support round-tripping between visual, AI, and code—no one-way streets.
Adopt standard frameworks like React, Angular, or Vue to avoid vendor lock-in.
Enable maintainability, so developers aren’t punished for scaling or customizing their applications.
Legacy RAD environments will fall behind if they don’t adapt to this new paradigm. Developers want the speed of AI and the clarity of visual design—without sacrificing control.
Tomi Engdahl says:
Triple-view development is not a gimmick—it’s a necessity. And it’s how we bridge the gap between classic RAD, Low Code and the intelligent, collaborative environments of tomorrow.
Tomi Engdahl says:
https://beyang.org/how-i-configure-vs-code-for-agentic-coding.html
Tomi Engdahl says:
Anthropic Claude Sonnet 4 and Claude Opus 4 are now generally available in GitHub Copilot
https://github.blog/changelog/2025-06-25-anthropic-claude-sonnet-4-and-claude-opus-4-are-now-generally-available-in-github-copilot/
Tomi Engdahl says:
5 Must-Have System Design Cheat Sheets for Interviews
https://dev.to/somadevtoo/5-must-have-system-design-cheat-sheets-for-interviews-a7c
Tomi Engdahl says:
https://www.xda-developers.com/linux-users-can-help-others-make-the-switch/
Tomi Engdahl says:
https://lemire.me/blog/2025/06/22/c26-will-include-compile-time-reflection-why-should-you-care/
Tomi Engdahl says:
https://www.productcompass.pm/p/multi-agent-research-system
I Copied the Multi-Agent Research System by Anthropic. No Coding!
A deep research n8n template with step-by step instructions. You can use those techniques for competitor analysis, outbound marketing, or lead generation.
Tomi Engdahl says:
https://wccftech.com/microsoft-founder-bill-gates-and-linux-creator-linus-torvalds-cross-paths-for-the-first-time-in-50-years/
Tomi Engdahl says:
Lessons from letting AI vibe code a landing page
An AI-powered experiment in copy, research and vibe coding — delivering faster results, fewer revisions and a few unexpected hurdles.
https://martech.org/lessons-from-letting-ai-vibe-code-a-landing-page/
Tomi Engdahl says:
https://dev.to/rohan_sharma/creating-a-chatbot-from-scratch-and-vibe-coding-the-ui-1bij
Tomi Engdahl says:
https://simonwillison.net/2025/Jun/21/my-first-open-source-ai-generated-library/
Tomi Engdahl says:
https://www.xda-developers.com/using-unbound-to-create-dns-server-instead-google-or-cloudflare/
Tomi Engdahl says:
https://lambgoat.com/news/48292/rick-rubin-calls-ai-vibe-coding-the-punk-rock-of-software-and-shares-new-digital-book/
Tomi Engdahl says:
https://devinterrupted.substack.com/p/no-more-coding-vibes-in-the-efficiency
Tomi Engdahl says:
https://madza.hashnode.dev/9-ai-productivity-tools-you-will-be-amazed-to-discover
Tomi Engdahl says:
https://probablydance.com/2025/06/19/revisiting-knuths-premature-optimization-paper/
Tomi Engdahl says:
https://simonwillison.net/2025/Jun/21/edit-is-now-open-source/
Tomi Engdahl says:
https://visualstudiomagazine.com/articles/2025/06/18/copilot-compared-advanced-ai-features-in-visual-studio-2022-vs-vs-code.aspx
Tomi Engdahl says:
https://www.pcgamer.com/software/operating-systems/denmark-is-switching-to-linux/
Tomi Engdahl says:
https://hackaday.com/2025/06/21/if-your-kernel-development-is-a-little-rusty/