Christian Findlay

And here it is!!! My new programming language Osprey! This has been a 20 year long dream, finally made possible with AI assistance. So stoked this is finally possible! 🎉 Stars and shares appreciated! ✨ ospreylang.dev

@jamonholmgren Congratulations 🥂

So. Uh. I did a thing. This is my game.

Should have made this video with Seedance. Would feel more natural

@burkeholland Hooks. The answer is pre and post tool use hooks

Manager introduces a new team member... "FYI - this person is a pathological liar. Just - just check everything they do and don't let them anywhere near production." Completely uncharted waters. We can fix this though. I believe this will be the year.

I've long preferred Claude Code over Codex or Gemini, because it seemed much more reliable, but couldn't explain why : now Bullshit Bench by @petergostev provides compelling numbers. It measures bullshit as "when given false premises disguised in jargon, will the model go with the flow (=bullshit) or push back (=truthful)" And Claude is leagues ahead ! Also, this objective of truthfulness is probably at odds with the Chatbot Arena emergent objective of "pleasant chat experience" ; but a model optimizing for the former will be more useful.

@alexanderOpalic Try this tmc-mcp.dev

How many agents can you run at the same time?

@CFDevelop The .fsx "pre-request script" feature makes this better than an .http file. #fsharp

@jordan_n_marr Awesome! Glad to hear it. And the latest version can convert .http files to nap files. If you have any issues, please let me know on GitHub

Did you check this out? Postman/Bruno alternative, CLI first, runs in vscode, supports C# and F#, fully open source Just install in vscode napperapi.dev

@ihtesham2005 This is exactly how everyone who knows what they’re doing are using AI right now.

🚨 Holy shit...A developer on GitHub just built a full development methodology for AI coding agents and it has 40.9K stars on GitHub. It's called Superpowers, and it completely changes how your AI agent writes code. Right now, most people fire up Claude Code or Codex and just… let it go. The agent guesses what you want, writes code before understanding the problem, skips tests, and produces spaghetti you have to babysit. Superpowers fixes all of that. Here's what happens when you install it: → Before writing a single line, the agent stops and brainstorms with you. It asks what you're actually trying to build, refines the spec through questions, and shows it to you in chunks short enough to read. → Once you approve the design, it creates an implementation plan so detailed that "an enthusiastic junior engineer with poor taste and no judgement" could follow it. → Then it launches subagent-driven development. Fresh subagents per task. Two-stage code review after each one (spec compliance, then code quality). The agent can run autonomously for hours without deviating from your plan. → It enforces true test-driven development. Write failing test → watch it fail → write minimal code → watch it pass → commit. It literally deletes code written before tests. → When tasks are done, it verifies everything, presents options (merge, PR, keep, discard), and cleans up. The philosophy is brutal: systematic over ad-hoc. Evidence over claims. Complexity reduction. Verify before declaring success. Works with Claude Code (plugin install), Codex, and OpenCode. This isn't a prompt template. It's an entire operating system for how AI agents should build software. 100% Opensource. MIT License.

@johncrickett My strong opinion: automated and manual testing are the two most critical skills for humans right now. All you gotta do is make sure it works and that the tests prove it works

I've been documenting my views on AI-assisted coding. I believe in strong opinions, weakly held, so I'd like thoughts, feedback, and challenges to the following opinions. On AI and AI-Assisted Coding: → The agent harness largely doesn't matter. The process should work with all of them. → Most AI-assisted coding processes are too complex. They clutter the context window with unnecessary MCP tools, skills, subagents, or content from the AGENTS file. → A small, tightly defined, and focused context window produces the best results. → LLMs do not reason, they do not think, they are not intelligent. They're simple text prediction engines. Treat them that way. → LLMs are non-deterministic. That doesn't matter as long as the process provides deterministic feedback: compiler warnings as errors, linting, testing, and verifiable acceptance criteria. → Don't get attached to the code. Be prepared to revert changes and retry with refinements to the context. → Fast feedback helps. Provide a way for an LLM to get feedback on its work. For example, tests, compilers and linters. → Coding standards and conventions remain useful. LLMs have been trained on code that follows common ones and to copy examples in their context. When your code align with those patterns, you get better results. On Software Development: → Work on small defined tasks. → Work with small batch sizes. → Do the simplest possible thing that meets the requirements. → Use TDD. → Make small atomic commits. → Work iteratively. → Refactor when needed. → Integrate continuously. → Trust, but verify. → Leverage tools. → Don't get attached to the code. What are your strong opinions on AI-assisted coding?

@jamonholmgren Massaging text was never the interesting or important part. Thinking about the thorny problems and trade offs is

@CFDevelop Yes, 100%

One of the many complex systems on my game is the radar/optical scanner system. As a combat helicopter game, each vehicle needs some way to know if it can see or detect another vehicle. The thing is, line of sight (LOS) calculations aren’t cheap. They have to detect collisions with rolling terrain, trees, I only have 16.67 ms at best to do all the work I need to do every frame, and I can’t be checking every single vehicle from every single other vehicle every frame. It would be terrible for performance. Instead, I check LOS for every vehicle from every vehicle in a round robin way — one per frame. So if there are 100 vehicles, that means it’ll take 99 frames for every vehicle to check every other vehicle (it doesn’t have to check itself). It’s not just that, though. It also needs to account for whether you can detect something immediately (it’s close and loud and obvious) or far away and you’re getting glimpses through the trees. Or you are at night, it’s foggy, and it’s an optical system - this degrades the system’s capability to detect something.

A robot in China just smashed some dishes started dancing instead of working 😂

Released `build_runner` 2.13, mostly performance improvements :) pub.dev/packages/build… It's faster across the board, but particularly for large projects: with these improvements plus an upcoming analyzer release, an incremental build on 10000 files that used to take 45s with 2.12 is now nearly 5x faster, just 9.5s. Full benchmark numbers are here: github.com/dart-lang/buil…

@artvepa @pvergadia This absolutely can and does happen. This is the one thing you have to be hyper vigilant about. Manual testing is the new automated testing

@CFDevelop @pvergadia The layering approach works until the AI-generated tests start encoding the bugs instead of catching them. Seen it happen test passes, behavior is wrong, nobody notices because the test was written from the same broken code it's covering.

🤯BREAKING: Alibaba just proved that AI Coding isn't taking your job, it's just writing the legacy code that will keep you employed fixing it for the next decade. 🤣 Passing a coding test once is easy. Maintaining that code for 8 months without it exploding? Apparently, it’s nearly impossible for AI. Alibaba tested 18 AI agents on 100 real codebases over 233-day cycles. They didn't just look for "quick fixes"—they looked for long-term survival. The results were a bloodbath: 75% of models broke previously working code during maintenance. Only Claude Opus 4.5/4.6 maintained a >50% zero-regression rate. Every other model accumulated technical debt that compounded until the codebase collapsed. We’ve been using "snapshot" benchmarks like HumanEval that only ask "Does it work right now?" The new SWE-CI benchmark asks: "Does it still work after 8 months of evolution?" Most AI agents are "Quick-Fix Artists." They write brittle code that passes tests today but becomes a maintenance nightmare tomorrow. They aren't building software; they're building a house of cards. The narrative just got honest: Most models can write code. Almost none can maintain it.

@wieslawsoltes @pvergadia It's probably not BS, it just doesn't tell the full story. You can use statistics to tell any story

@CFDevelop @pvergadia That research is BS, you can migrate and improve legacy code it takes some work but definitely possible and it’s maintainable in long run, speaking from experience 😎

This is one is very nice

@Pranit And the unethical approach would be shocking if there was a modicum of an objective way to measure how many tokens we’re getting each month to start with…

Anthropic just pulled the oldest trick in SaaS pricing. I pay $200/mo for Claude Max. My limits have been noticeably worse this past week. Now they announce 2x off-peak usage for two weeks. Sounds generous. But here’s what actually happens: limits quietly drop, a temporary 2x makes the reduced limit feel normal, the promo ends, and you’re left at a baseline lower than where you started. You just didn’t notice the downgrade because the 2x absorbed the transition. These AI plans are massively subsidized. The raw compute behind a heavy user costs multiples of the subscription price. Every move like this is the subsidy quietly correcting. Very sneaky, Anthropic.

@rtfeldman @AdedamolaFunso The studies support this

In the past 3 years, I haven't noticed any uptick in release speed for software I use. If productivity is increasing, I can't tell as an end user. I have noticed decreases in uptime, increases in bugs, and a HUGE increase in people bragging about how many PRs per day they land.