6. change thread title color

5. improve mobile UI

#idea #EnglishCoach My roommate is preparing for her IELTS exam, so I want to build a tool that can give her tailored feedback on her practice. Since it's for her daily study, I'm thinking of a quick, free CLI app that reviews her writing based on her actual notes. [Planned Features] * AI-driven evaluation & writing feedback * Note-based personalization (trained/prompted on her IELTS notes) * Free CLI interface for daily practice * Band-score improvement tips & grammar checks

#idea #dailybit [Future Features] 1. Allow editing existing threads 2. Add title input when creating a thread 3. Bold styling support in editor 4. Expand the text input area

#aichatbot At my last company, I built a chatbot that recommended products from our online fashion store (Gemini + Milvus RAG)It ran on our in-house data: product info like category, color, material and season, plus customer reviews and sales numbers.After I left, I couldn't use that data anymore, so I rebuilt it on Amazon's public review dataset (8,500 clothing items).

Next up: - Security fixes before going public (admin API auth, input length limits) - Deploying on a free server so anyone can try it 🔗 github.com/summerr0-0/chatbot_project

The new data needed changes: - Korean values became English, prices became USD - Amazon has no clean color/material/season fields, so I extract them from product text And for a portfolio: - Balanced picks across categories, with Amazon's own brands standing in as "our brand" - A live monitoring page that shows every reasoning step - Rate limits to cap API costs

Key features: - 3-way hybrid search: mood, specs and keywords - An LLM reads each question first, pulls out filters (category, color, budget) and asks back if it's vague - Filters run inside the DB, so jeans never return a jacket - Ranking boost for popular items and our own brand - Reviews summarized by AI and searched as specs

#AnkiWatch #Idea I use Anki for vocab and studying, but since starting a part-time job, I barely look at my phone. My watch, though? I wear it every day and check it all the time. So what if Anki showed up on my watch? Tag cards "watch-sync" in Anki → the app pulls them from AnkiWeb → a word pops up on my wrist every 15/30/60 min (or N random times a day). Keep your wrist raised to see the meaning, then tap "Next" or "Done for today." Thinking of building it like this.

When multiple requests come in for the exact same long URL, you face a distinct architectural trade-off: Deduplication: Save the mapping only once. This significantly optimizes storage capacity, but introduces an extra read check before every single write, causing minor write latency.

Therefore, tracking-focused companies must use 302 (Temporary Redirect). It tells the browser never to cache the link, forcing every single click to hit your servers for precise analytical tracking. To mitigate the heavy traffic load this introduces, we isolate our read/write paths by routing requests through a Redis cache cluster and distributed database read replicas.

There are two primary HTTP redirection statuses to consider: 301 and 302. 301 (Permanent Redirect) triggers aggressive browser caching. This is incredibly fast and removes almost all overhead from your infrastructure. However, because subsequent clicks bypass your servers entirely, marketers lose the ability to track ongoing user engagement data.

To solve this, we use the Twitter Snowflake algorithm. It generates highly unique, time-sorted 64-bit IDs directly in the application server's memory, combining absolute speed with optimized database indexing.

But for a large system, this approach fails. Centralized auto-increment IDs create a critical write bottleneck under heavy concurrent loads. On the other hand, using random UUIDs degrades database performance because unordered strings destroy sequential index efficiency, leading to random disk I/O.

For a small system (under 10,000 writes/day), keep it simple. Use the database's Auto-Increment ID and convert that integer to Base62 to create the short URL alias. Database sequential IDs are excellent because they are clustered indexes, making them highly optimized for fast disk lookups.

#system_design Let me work through how to design a URL shortener service. It is used for marketing tracking (to see where users click) and preventing link breakage when URLs are too long. When designing this, we must always clarify the system constraints first: - Is this specifically for marketing tracking? -What is the scale (how many write/read requests per day)?

Unique Link Generation: Issue a completely unique short link for every single request. This is ideal for tracking separate marketing campaigns for the same product, but it risks explosive database growth. To prevent this storage bloat, implementing a strict TTL (Time-To-Live) eviction policy to automatically purge stale links is mandatory.