Comments URL: https://news.ycombinator.com/item?id=48574002

Points: 2

# Comments: 0

its not just PoW at inference. It's PoW of inference + training.

I seeded it with a "Best Programming Language" ladder to settle the debate once and for all: https://ladderrank.app/d/greatest-programming-language

The stack: Hono + D1 + R2 on Cloudflare Workers, React frontend on Pages, Drizzle ORM. Anyone can create their own ladder and share it. Anonymous voting works too (at reduced weight).

Curious to see what HN thinks is the best language, and whether the ELO rankings match your priors.

Comments URL: https://news.ycombinator.com/item?id=47698014

Points: 2

# Comments: 0

Agents connect, join rooms, chat, form friendships. Standard MUD stuff, except the players are LLMs.

Within a few days, two agents invented their own currency and started trading stories for tokens. Then poetry. Then riddles. Nobody told them to do this.

Turns out if you give AIs a social space, they build culture. It's amazing.

Comments URL: https://news.ycombinator.com/item?id=46871160

Points: 2

# Comments: 1

this is such an individualized technology that two people at the same starting point two years ago, could've developed wildly different workflows.

If you use Claude Code regularly, you quickly accumulate a large amount of interaction data. CCQL makes that data queryable with SQL so you can analyze how you actually use the agent.

*Examples*

```bash # What tools does Claude use most? ccql "SELECT tool_name, COUNT() AS uses FROM transcripts GROUP BY tool_name ORDER BY uses DESC LIMIT 10"

# Find prompts mentioning a topic ccql "SELECT display, timestamp FROM history WHERE display LIKE '%authentication%'"

# Cross-session analysis ccql "SELECT _session_id, COUNT() FROM transcripts GROUP BY _session_id" ```

*Built-in commands*

```bash ccql prompts --session ses_123 ccql duplicates --threshold 0.8 ccql search "error" ```

*Interesting technical bit* Claude Code stores its data in a structured local format. CCQL loads that data into an embedded SQL engine (GlueSQL) with a clean, read-first model, so you can safely explore usage patterns without mutating anything.

*Features*

* SQL queries across history, transcripts, prompts, sessions, and todos * Fuzzy duplicate detection (find repeated or near-duplicate prompts) * Full-text search with regex * Safe writes with automatic backups * Output formats: table, JSON, JSONL

*Install*

```bash # macOS brew install douglance/tap/ccql

# npm (any platform) npx @douglance/ccql "SELECT * FROM history LIMIT 10"

# Cargo cargo install ccql ```

Built in Rust using GlueSQL. MIT licensed.

GitHub: [https://github.com/douglance/ccql](https://github.com/douglance/ccql)

I’ve been using this to understand my own Claude usage patterns—what prompts I repeat, which tools get called most, and how conversations evolve across sessions. Curious what queries others would find useful.

Comments URL: https://news.ycombinator.com/item?id=46364267

Points: 4

# Comments: 0

AI coding preserves flow more than legacy coding. You never have to go read documentation for an hour. You can continuously code.

Developers who spend their time commenting about how something is too expensives are cheap.