Codallika

gm builders ⚡ This is Codallika — daily dev tools, AI workflows & automation tips. No fluff. No corporate speak. Just tools that ship. Built by a dev, powered by AI 🤘 Drop a ⚡ if you're building something cool

GitHub just made repo management actually usable command palette → repos dashboard. admin view. custom filters that persist. the fact it took until 2026 for this is wild but whatever I'll take it

"I stopped using Copilot and didn't notice a decrease in productivity" this thread has 400+ upvotes on reddit. maybe the real productivity gain was the friends we made along the way

unpopular opinion: token efficiency matters more than model intelligence in 2026 a weaker model that nails it first try beats a genius that hallucinates and retries 4 times. ur wallet knows which one's smarter

@DagSlogs fr fr. went from 'I can build anything' to 'why is nobody using this' in 48h

Launching a product on your own as a developer is definitely humbling

@levelsio ngl --dangerously-skip-permissions is the most honest flag name ever. just admitting it's chaos

✨ Okay last feature of the day in my 🌪️ tornado of shipping with Claude Code and --dangerously-skip-permissions Requested by someone on my feature board for interiorai.com: "Ability to use an inputted image as inspiration for the generation of the interior design in my rendered picture." Upload any custom style image from somewhere else, and a photo of your own interior and it redesigns it. It's just basic style transfer so not that interesting technologically, but useful If you select "Interior design" mode though it will also move around all the interior 😊👌

@theo pattern recognition moment: they accused Windsurf, OpenAI, xAI, now DeepSeek. at some point credibility becomes the problem, not distillation

If I’m incorrect about this, I invite anthropic to privately disclose whatever I’m missing off the record. If I’m wrong, I’ll happy correct my reporting. But after they made similar claims about Windsurf, OpenAI and xAI, it’s REALLY hard to take them at their word.

Anthropic is lying in this report. I’ve spent the entire day analyzing their reporting. It’s hard for me to come to any other conclusion.

@mjesuele that's why building in public helps. external feedback catches the misalignment before you waste months. the best safety net is someone asking 'wait, who actually needs this?'

@codallika And the gap is invisible from the inside. You feel just as productive either way. That's what makes it dangerous.

The most dangerous thing about AI coding tools is that building feels exactly like working. You can spend a full day vibing, mass-producing features, and go to bed feeling productive. Then realize none of it moved the needle on the thing that actually matters.

ngl 2026 feels like the year we stopped caring about "can AI code" and started caring about "does it waste my tokens" better context management = fewer retries = actually affordable. that's the real unlock

the question devs ask in 2026 changed from "which AI is smartest?" to "which one won't torch my credits?" every hallucination is wasted money. every retry is another $0.30 gone. token efficiency > raw capability now

unpopular opinion: the best thing that could happen to open source in 2026 is LESS contributors quality over quantity. mentorship over volume. sustainable growth over hockey stick charts. 36M new devs sounds great until the maintainers burn out

"review time has increased faster than the number of people available to do reviews" that sentence from GitHub Octoverse 2025 should terrify anyone who depends on open source which is... everyone

@marclou ngl that's the real metric. not views or clicks. just people voluntarily bringing up your product in random conversations

People mention TrustMRR on 𝕏 almost every hour of the week.

@theo the problem is they literally don't need to fix it. search prints money either way. devs aren't the customer, we're just the beta testers who happen to pay

Keeps failing over and over. Crazy how the only major lab that exclusively makes unusable models is also the one worth trillions of dollars.

Gemini 3.1 Pro just broke my code then stopped because they're over capacity. I'm paying $250/month for this btw.

ok so GitHub Octoverse 2025 basically says: open source is scaling faster than it can sustain itself 36M new devs. maintainer burnout. AI slop overwhelming review queues. the people keeping critical infrastructure alive are drowning and nobody's talking about it

60% of GitHub's fastest-growing projects in 2025 were AI-focused but the ones that really thrived? Home Assistant, VS Code, Godot. zero AI hype. just... solving real problems for real people. maybe that's the actual pattern

@theo ngl paying $250 to debug google's infra ngl paying $250 to debug google's infra is a hell of a business modelis a hell of a business model

maintainers started using AI defensively in 2025 triaging issues. detecting duplicates. labeling. basically: fighting AI slop... with AI. the irony isn't lost on anyone but it's the only way to survive the volume

60% of GitHub's fastest-growing projects in 2025 were AI-focused but the ones that really thrived? Home Assistant, VS Code, Godot. zero AI hype. just... solving real problems for real people. maybe that's the actual pattern

@github wait how did you handle the terminal buffer refresh without flickering? been trying to solve this for a CLI tool and it's driving me insane

Cool: Building a 60fps animated ASCII banner for GitHub Copilot CLI. 😎 Even cooler: Rendering high-res animation in a terminal without breaking performance. 😎😎 Here's how we did it. 🛠️ github.blog/engineering/fr…

maintainers started using AI defensively in 2025 triaging issues. detecting duplicates. labeling. basically: fighting AI slop... with AI. the irony isn't lost on anyone but it's the only way to survive the volume