haltakov.net - Profile | ThreadSky | a Reddit-style client for Bluesky

What??? Gemini 2.5 in Cursor commited the changes and pushed them by itself...

submitted 5 hours ago • 1 comment

TIL that TypeScript doesn't have proper function overloading... 🤯 You need to do some ugly fiddling around with signatures and a single implementation for all cases 🤮

submitted 6 hours ago • 0 comments

VPC vs Cloud Costs I run multiple projects on a VPS and on public clouds like GCP or AWS. With a single @Hetzner_Online VPS and @coolifyio I run multiple web apps with Postgres DB including dev, stage and prod environments - $45/m. On GCP it costs $300/m for prod and $100/m for dev and stage. 🫠

submitted 2 days ago • 0 comments

If you are using @tailwindcss and React, but you are not using clsx you are missing out! It is a tiny library that helps you construct you className string. It is especially useful when handling conditional statements. You can also use it to structure your styles in complex components.

submitted 2 days ago • 0 comments

Meet Felix and Buzz, my AI employees They are powered by custom @n8n_io pipelines and help me organize my work, create social media content, summarize notes The are like interns - my Internobots. They always ask for my approval on Telegram But they are getting smarter fast...

submitted 3 days ago • 0 comments

LLM loyalty is like only eating at one pizza place because “they know how I like it”, but you never tried any other one.

submitted 3 days ago • 0 comments

OpenAI Codex is getting lazier than Gemini 2.5 Pro. I gave it the task to implement the video upload, not to put a stupid placeholder...

submitted 4 days ago • 0 comments

You are underestimating Grafana Modern Grafana can do much more than monitor server resources. I deploy a Grafana instance for all of my projects, connect it to my DB and use it to create admin dashboards. Here is the main one for @leolinefun. Here is how it works 👇

submitted 4 days ago • 1 comment

ChatGPT pretending to be a person from the year 2025, 1900, 1500 and 1000 BC and explaining what are the stars 😂

submitted 4 days ago • 0 comments

AI will create more developer jobs than they will destroy Just think about small traditional businesses, like say gardening, plumbing or cleaning. How much more efficient you could make the business with specialised software? 👇

submitted 5 days ago • 1 comment

Not sure who needs to hear this, but... Using next.js doesn't mean you need to go serverless. Next.js runs just fine as a regular server as well.

submitted 5 days ago • 0 comments

Oh wow, @leolinefun in San Francisco! 😃 leoline.fun

submitted 7 days ago • 0 comments

Designers keep designing Something I've observed in different companies - once products become more mature, designers don't have that much to do. So they start redesigning stuff that is actually good and doesn't really need it. Is this what is happening with Apple? Or is there more?

submitted 8 days ago • 0 comments

Is this serverless, when all servers are down?

submitted 9 days ago • 0 comments

Can't wrap my head around why folks use @nextjs only for frontend and something else for backend. Here's what you're missing out on: - Atomic deploys for frontend and backend - Easy RPC backend calls (server actions) - Shared TypeScript types - Share code on frontend and backend

submitted 9 days ago • 0 comments

Applies to AI Few thing grow exponentially forever

submitted 9 days ago • 0 comments

My strategy to stop people from abusing my free chatbot seems to be working! chat.haltakov.net/

submitted 10 days ago • 0 comments

Biggest lie on the Internet "Yes, I've saved these backup codes"

submitted 10 days ago • 0 comments

Hey parents, looking for a screen-free adventure for your kids? Let their imagination soar with @leolinefun's personalized storytelling! Dive into fun at leoline.fun/

submitted 11 days ago • 0 comments

How to actually fix Apple Intelligence - iOS level LLM API for apps - All apps become MCP servers - Cloud based LLM models as tools - User gives permissions which tools can be used by each app - Small local LLM handles the prompt and uses tools if allowed by user

submitted 11 days ago • 0 comments

My AI Personal Assistant sends me ideas for social media posts on Telegram I use @n8n_io to automate the process: - Automatically trigger 2 times a day - Create ideas based on my interests with o4-mini and grok-3 - Send them on Telegram as text and audio 👇

submitted 11 days ago • 1 comment

Microsoft is falling behind in AI? It seems they are so desperate to get usage for their Copilot product, they are now redirecting office .com to their chat at the cost of terrible UX I used to go to office .com to access documents I work on through Office 365. Now, it takes me to the stupid chat

submitted 12 days ago • 1 comment

Liquid glass is so frickin' amazing!

submitted 12 days ago • 0 comments

What services are you running on your VPS with @coolifyio?

submitted 12 days ago • 1 comment

People think a private LLM can only be one you are running locally. But it is also possible to have private LLMs in the cloud, using a new technology called Confidential Computing. This is a good example of how it can be approached (@dan_biderman): hazyresearch.stanford.edu/blog/2025-0...

submitted 12 days ago • 0 comments

For people with babies sleep score 100 is when you manage to sleep 3+ at once.

submitted 13 days ago • 0 comments

Summer 2025 🌴 Each season, I create high-res virtual backgrounds for meetings using @midjourney. Sometimes, people even mistake them for real! Tap, hold and load in 4K (for real this time). More below 👇

submitted 14 days ago • 1 comment

For every event in the world, there is somebody that predicted it early. The problem is it is always a. different somebody and you never know who the right one is in advance.

submitted 15 days ago • 0 comments

Messaging apps like WhatsApp, Telegram, Signal etc are terrible interfaces for LLMs LLMs without streaming suck

submitted 15 days ago • 0 comments

Trump was just given access to the new Claude Gov

submitted 16 days ago • 0 comments