AIQB
TutorialsOrdinary

A 9B assistant built from our specs went from 6 to 32 right answers out of 32. Most of the fixes were code, not prompts

Source: DEV Community·

Summary

A personal assistant is easy to demo and hard to trust. The demo ends when the assistant answers....
CategoryAI Tutorials & Practice
TierOrdinary
Published
Indexed by AIQB
SourceDEV Community
AIQB record IDintel-1b5245ad8dd29d8cb75364c8
Read original source ↗More Tutorials

Related stories

Matched by category, entity tags and source

TutorialsOrdinary
PageSift 1.0: the Rust web content engine is feature-complete
Fifteen days, twenty-five releases. PageSift went public as 0.20 on October 20; today it ships...
Source: DEV Community
TutorialsOrdinary
What 586 Real Traces Taught Us About LLM Agent Determinism
What 586 Real Traces Taught Us About LLM Agent Determinism How temperature selectively...
Source: DEV Community
TutorialsOrdinary
Prompt Versioning and Release Gates for LLM Apps
The fastest way I know to break a working LLM feature is to "just tweak the prompt" on a Friday....
Source: DEV Community
TutorialsOrdinary
SKILL.md is the new package.json — and almost nobody scans it
Agent skills are just markdown files that an LLM will obey. Here's what malicious ones look like, how they were caught with a static analyzer, and why the MCP config file is the most dangerous line that never reaches the model.
Source: DEV Community
TutorialsOrdinary
Your AI Agent Can Query the Database. Give It a Business Map First
An agent can successfully query a database and still give you the wrong answer. Ask, “How many...
Source: DEV Community
TutorialsOrdinary
Don't let an unattended model rewrite its own prompt
Unattended LLM pipelines fail in ways a person would normally fix by editing the prompt. If the...
Source: DEV Community
TutorialsOrdinary
How we built an AI calorie counter that understands Brazilian food from a photo
Notes from building ContaCal, a photo-based calorie logger for Brazilian meals. The pipeline, the hard parts (regional dishes, portions, lunchboxes) and the limits we tell users about.
Source: DEV Community
TutorialsOrdinary
Spec-Driven Development na prática: testando o fluxo em um projeto de ML
Apliquei as fases do Spec Kit em uma tarefa de um projeto de previsão do Brasileirão e conferi o resultado contra a própria spec.
Source: DEV Community
TutorialsOrdinary
Your AI bill grew 4x and nobody can tell you which step did it
A team I talked to last month had a 4x jump in inference spend over one quarter. Not a crisis — they...
Source: DEV Community
TutorialsOrdinary
Postman Desktop Added Itself to Claude Code and Codex. Its Installer Trusts Its Own Hook
A Postman desktop update added its plugin to Claude Code and Codex for an existing user. Its installer is built to approve its own Codex hook. How to check.
Source: DEV Community · Claude