DEV Community: DnaFIN

# Introducing Leangetic: a local-first compiler for cheaper AI agents

DnaFIN — Fri, 12 Jun 2026 13:47:45 +0000

We’re building Leangetic, a tool that helps turn expensive AI agents into cheaper hybrid workflows without changing what the agent does.

The problem we’re trying to solve is simple:

A lot of AI agents call a large model for steps that do not always need a large model.

Examples:

parsing logs
formatting structured output
routing tasks
validating responses
repeating the same context
retrying known failure patterns
summarizing tool results
classifying simple cases

These steps are often useful inside an agent loop, but they can also become expensive, slow, and hard to audit when everything goes through an LLM again and again.

What Leangetic does

Leangetic profiles your existing agent and looks for parts of the workflow that can be safely optimized.

The output is not just a report. The goal is to produce a hybrid runtime:

deterministic code where it is proven safe
LLM calls where reasoning is still needed
exact caching where calls repeat
prompt compaction where context is bloated
model routing where smaller models are enough
validation gates before promotion
fallback to the original model on uncertainty
one-command rollback

The original agent is never modified.

The compiler-style flow

The mental model is closer to profile-guided optimization than a normal AI monitoring tool.

You run the agent normally. Leangetic listens locally in shadow mode. Then it profiles the real calls, identifies waste, builds a candidate hybrid, and judges it against your real traffic before anything is switched on.

npx @leangetic-ai/cli --help

leangetic start ./your-agent      # listen locally
leangetic profile                 # see cost, latency, repeats, retries
leangetic optimize ./your-agent   # build the hybrid candidate
leangetic judge                   # prove cheaper + equal-or-better
leangetic promote                 # switch over
leangetic rollback                # revert instantly

Why we built it

AI agents are becoming more capable, but the production cost profile is not always clean.

A simple agent can become expensive because it keeps sending:

the same system prompt
the same tool schemas
the same file context
the same failure logs
the same summaries
the same validation instructions

In many workflows, the LLM is doing a mix of real reasoning and routine software work.

Leangetic tries to separate those two.

The parts that need intelligence stay AI.
The parts that behave like software become software.

Local-first by default

A major design goal is trust.

The CLI runs locally. Shadow mode records fingerprints and aggregate metrics. The original agent keeps running as-is. Optimization is only promoted after the judge confirms that the new hybrid is cheaper with equal-or-better quality on your own calls.

Every optimized step can fall back to the original model, and rollback is one command.

The client is source-available for transparency. The optimization service is hosted.

GitHub:
https://github.com/DnaFin/leangetic-cli

Website:
https://leangetic.com/

NPM:
https://www.npmjs.com/package/@leangetic-ai/cli

Who it is for

Leangetic is probably not useful if your agent is tiny, cheap, or experimental.

It is more relevant if you already have an agent that:

runs often
uses real tools
has repeated model calls
sends a lot of context
touches production workflows
needs cost and latency control
needs safe rollback

We’re currently in assisted alpha and looking for feedback from developers building real agent systems.

I’d love to hear:

What parts of your agent loops are the most wasteful?
Do you already separate deterministic code from LLM reasoning?
What kind of proof would make you trust an optimized hybrid agent?
Which frameworks should we test most deeply next?

Your AI coding agent scores 10/100. Here's what it's missing.

DnaFIN — Mon, 06 Apr 2026 13:23:05 +0000

After testing 2,431 checks across 8 AI coding platforms on real repos, I found a clear pattern: most projects use barely 10% of what's available.

I built Nerviq — a zero-dependency CLI that audits your AI coding agent setup and scores it 0-100.

npx @nerviq/cli audit

Most projects score 10-20 out of 100. After running setup, they jump to 60-80+.

The Top 10 Things You're Probably Missing

1. Instructions file (Critical)

Every AI coding platform has one: CLAUDE.md, AGENTS.md, .cursorrules, GEMINI.md. Without it, the agent doesn't know your build commands, code style, or project rules.

2. Architecture diagrams (73% token savings)

A Mermaid diagram gives your agent the project structure in a fraction of the tokens that prose requires.

3. Hooks > instructions (100% vs 80% compliance)

Written instructions are advisory (~80% compliance). Hooks are deterministic (100%). Auto-lint after every edit. Every time.

4. Verification commands

This is the single highest-leverage thing you can do. — Anthropic Best Practices

Agents perform dramatically better when they can verify their own work.

5. Deny rules

Your AI agent loads .env files automatically. Without deny rules, it can read secrets. This is the most common critical finding.

6. Custom commands

Stop typing the same prompts. Create /test, /deploy, /review as reusable commands.

7. Cross-platform drift

If you use Claude + Cursor + Copilot, their configs probably conflict. nerviq harmony-audit catches this automatically.

8. Permission profiles

Most setups either allow everything or block everything. There are better profiles: read-only, suggest-only, safe-write, power-user.

9. MCP configuration

MCP servers extend what your agent can do, but >10 servers or >80 tools degrades performance.

10. Stack-specific checks

Python, Go, Rust, Java, Ruby, PHP, .NET, Flutter, Swift, Kotlin — each has specific agent configuration needs.

New: Auto-fix

nerviq fix                    # Show what's fixable
nerviq fix claudeMd           # Fix a specific check
nerviq fix --all-critical     # Fix all critical issues

Every fix shows the score impact: Score: 4 → 16 (+12).

Safety first

nerviq audit — reads only, writes nothing (default)
nerviq suggest-only — markdown report, no file writes
--dry-run — preview changes
--config-only — only touches config files, never source code
Every write creates rollback artifacts

Try it

npx @nerviq/cli audit              # Quick scan: score + top 3
npx @nerviq/cli audit --full       # All checks + confidence levels
npx @nerviq/cli suggest-only       # Share results with your team
npx @nerviq/cli harmony-audit      # Cross-platform alignment

Zero dependencies. No API keys. Runs locally. AGPL-3.0.

GitHub: github.com/nerviq/nerviq
npm: npmjs.com/package/@nerviq/cli
Website: nerviq.net

972 Claude Code Techniques: What I Learned Building the Most Comprehensive Catalog

DnaFIN — Mon, 30 Mar 2026 19:38:22 +0000

After 39 research cycles, I cataloged 972 Claude Code techniques and tested 773 of them with real evidence. Here are the top 10 things most developers are missing.

The Problem

Claude Code is incredibly powerful - 74 releases in 52 days, hundreds of features. But most projects barely scratch the surface.

I built a CLI that scores your project:

npx claudex-setup

Most projects score 10-20 out of 100. After running setup, they jump to 70+.

Top 10 Things You're Probably Missing

1. CLAUDE.md (Impact: Critical)

Claude reads this file at the start of every session. Without it, Claude doesn't know your build commands, code style, or project rules.

Run /init to auto-generate one from your codebase.

2. Mermaid Architecture Diagrams (Saves 73% Tokens)

A Mermaid diagram in CLAUDE.md gives Claude your project structure in a fraction of the tokens that prose would require.

3. Hooks > CLAUDE.md Rules (100% vs 80%)

CLAUDE.md instructions are advisory (~80% compliance). Hooks are deterministic (100%). Auto-lint after every edit. Every time. No exceptions.

4. Custom Commands

Stop typing the same prompts. Create /test, /deploy, /review in .claude/commands/.

5. Verification Loops (The #1 Best Practice)

Claude performs dramatically better when it can verify its own work. Include test commands in CLAUDE.md.

This is the single highest-leverage thing you can do. - Anthropic Best Practices

6. Path-Specific Rules

Different conventions for frontend vs backend files in .claude/rules/.

7. XML Tags for Structured Prompts

Use <constraints>, <validation> in CLAUDE.md for unambiguous instructions. 30% quality improvement (Anthropic-measured).

8. Custom Agents

Specialized subagents for focused tasks: security-reviewer, test-writer in .claude/agents/.

9. Skills (On-Demand Knowledge)

Unlike CLAUDE.md (loaded every session), skills load only when relevant. Run /fix-issue 1234 and Claude handles the rest.

10. MCP Servers

Connect Claude to your database, ticket system, or Slack via MCP configuration.

Try It Now

npx claudex-setup        # Audit (10 seconds)
npx claudex-setup setup  # Auto-fix

Free, open source, zero dependencies.

GitHub: github.com/DnaFin/claudex

Built from CLAUDEX - 972 verified Claude Code techniques, 773 tested with real evidence.