Skip to content

Commit 74ebdb4

Browse files
authored
Merge pull request #2 from inmve/add-new-techniques-from-ai-stack
Added 6 new techniques
2 parents 63b7048 + 23401e9 commit 74ebdb4

File tree

5 files changed

+102
-10
lines changed

5 files changed

+102
-10
lines changed

README-de.md

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,5 @@
1-
> **Aktive Entwicklung** — Aktualisiert am 5. September 2025
2-
> Neueste: Technik "Offene Fragen stellen, keine suggestiven" hinzugefügt
1+
> **Aktive Entwicklung** — Aktualisiert am 11. September 2025
2+
> Neueste: 6 neue Techniken aus dem ai_stack-Projekt hinzugefügt
33
> [Alle Updates ansehen →](CHANGELOG.md)
44
55
**Verfügbare Sprachen:** [English](README.md) | [Español](README-es.md) | [Deutsch](README-de.md) | [Français](README-fr.md) | [日本語](README-ja.md)

README-es.md

Lines changed: 52 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,5 @@
1-
> **Desarrollo Activo** — Actualizado 5 de septiembre de 2025
2-
> Último: Añadida técnica "Haz Preguntas Abiertas, No Dirigidas"
1+
> **Desarrollo Activo** — Actualizado 11 de septiembre de 2025
2+
> Último: Añadidas 6 nuevas técnicas del proyecto ai_stack
33
> [Ver todas las actualizaciones →](CHANGELOG.md)
44
55
**Idiomas disponibles:** [English](README.md) | [Español](README-es.md) | [Deutsch](README-de.md) | [Français](README-fr.md) | [日本語](README-ja.md)
@@ -224,6 +224,16 @@ Delega tareas aburridas, sistemáticas y que consumen tiempo a la IA - desde peq
224224
> "El mejor ejemplo que he encontrado para el agente fue migrar una aplicación enorme de una librería de UI a otra. No es trabajo difícil, pero toma una cantidad enorme de tiempo y es completamente desinteresante."
225225
> — Traducido por Claude
226226
227+
### Proporcionar Contexto para Nuevas Librerías
228+
229+
Cuando uses librerías fuera de los datos de entrenamiento de la IA, proporcionale ejemplos recientes y documentación para enseñarle cómo funciona la librería.
230+
231+
> "LLMs can still help you work with libraries that exist outside their training data, but you need to put in more work—you'll need to feed them recent examples of how those libraries should be used as part of your prompt."
232+
> [Simon Willison](https://simonwillison.net/2025/Mar/11/using-llms-for-code/#:~:text=LLMs%20can%20still%20help%20you%20work%20with%20libraries)
233+
234+
> "Los LLMs aún pueden ayudarte a trabajar con librerías que existen fuera de sus datos de entrenamiento, pero necesitas hacer más trabajo—necesitarás alimentarlos con ejemplos recientes de cómo esas librerías deben ser usadas como parte de tu prompt."
235+
> — Traducido por Claude
236+
227237
### Tratar la IA como un Interno Digital
228238

229239
Da a la IA instrucciones extremadamente precisas y detalladas como lo harías con un interno - proporciona firmas de función exactas y deja que maneje la implementación.
@@ -294,6 +304,36 @@ Haz que la IA escriba pruebas comprehensivas basadas en el comportamiento espera
294304
> "Pide a Claude que escriba pruebas basadas en pares de entrada/salida esperados. Sé explícito sobre el hecho de que estás haciendo desarrollo basado en pruebas para que evite crear implementaciones simuladas, incluso para funcionalidad que aún no existe en la base de código. Dile a Claude que ejecute las pruebas y confirme que fallan. Pide a Claude que confirme las pruebas cuando estés satisfecho con ellas. Pide a Claude que escriba código que pase las pruebas, instruyéndole que no modifique las pruebas."
295305
> — Traducido por Claude
296306
307+
### Pedir al Agente que Revise su Propio Código
308+
309+
Haz que la IA realice una revisión de código de su propio trabajo antes de la revisión humana para descubrir problemas y mejoras.
310+
311+
> "Asking the agent to perform a code review on its own work is surprisingly fruitful."
312+
> [Chris Dzombak](https://www.dzombak.com/blog/2025/08/getting-good-results-from-claude-code/#:~:text=Asking%20the%20agent%20to%20perform%20a%20code%20review)
313+
314+
> "Pedir al agente que realice una revisión de código de su propio trabajo es sorprendentemente fructífero."
315+
> — Traducido por Claude
316+
317+
### Uno Escribe, Otro Revisa
318+
319+
Haz que un agente escriba código, luego usa un agente nuevo para revisar y encontrar problemas.
320+
321+
> "Use Claude to write code. Run `/clear` or start a second Claude in another terminal. Have the second Claude review the first Claude's work. Start another Claude (or `/clear` again) to read both the code and review feedback. Have this Claude edit the code based on the feedback. This separation often yields better results than having a single Claude handle everything."
322+
> [Anthropic](https://www.anthropic.com/engineering/claude-code-best-practices#:~:text=Have%20one%20Claude%20write%20code)
323+
324+
> "Usa Claude para escribir código. Ejecuta `/clear` o inicia un segundo Claude en otra terminal. Haz que el segundo Claude revise el trabajo del primer Claude. Inicia otro Claude (o `/clear` nuevamente) para leer tanto el código como los comentarios de la revisión. Haz que este Claude edite el código basándose en los comentarios. Esta separación a menudo produce mejores resultados que hacer que un solo Claude maneje todo."
325+
> — Traducido por Claude
326+
327+
### Editar Código en el Diff
328+
329+
Revisa los cambios en la vista de diff y escribe correcciones directamente en el diff antes de confirmar.
330+
331+
> "I manually review all AI-written code and test cases. I'll add test cases for anything I think is missing or needs improvement, either manually or by asking the LLM to write those cases (which I then review)."
332+
> [Chris Dzombak](https://www.dzombak.com/blog/2025/08/getting-good-results-from-claude-code/#:~:text=I%20manually%20review%20all)
333+
334+
> "Reviso manualmente todo el código y casos de prueba escritos por IA. Añadiré casos de prueba para cualquier cosa que creo que falta o necesita mejora, ya sea manualmente o pidiendo al LLM que escriba esos casos (que luego reviso)."
335+
> — Traducido por Claude
336+
297337
## Técnicas Transversales
298338

299339
### Ejecutar Múltiples Agentes en Paralelo
@@ -342,6 +382,16 @@ Crea herramientas que respondan rápidamente, proporcionen mensajes de error cla
342382
> "Las herramientas necesitan ser rápidas. Mientras más rápido respondan (y menos salida inútil produzcan) mejor. Los crashes son tolerables; los cuelgues son problemáticos. ¡Las herramientas necesitan ser amigables para el usuario! Las herramientas deben informar claramente a los agentes sobre mal uso o errores para asegurar progreso hacia adelante. Las herramientas necesitan estar protegidas contra un mono del caos LLM usándolas completamente mal. ¡No existe tal cosa como error de usuario o comportamiento indefinido!"
343383
> — Traducido por Claude
344384
385+
### Usar Énfasis Fuerte en Prompts
386+
387+
Usa IMPORTANTE, NUNCA, SIEMPRE liberalmente en prompts para dirigir a la IA lejos de errores comunes - sigue siendo el enfoque más efectivo.
388+
389+
> "Unfortunately CC is no better when it comes to asking the model to not do something. IMPORTANT, VERY IMPORTANT, NEVER and ALWAYS seem to be the best way to steer the model away from landmines. I expect the models to get more steerable in the future and avoid this ugliness. But for now, CC uses this liberally, and so should you."
390+
> [Vivek (MinusX AI Team)](https://minusx.ai/blog/decoding-claude-code/#:~:text=Unfortunately%20CC%20is%20no%20better)
391+
392+
> "Desafortunadamente CC no es mejor cuando se trata de pedirle al modelo que no haga algo. IMPORTANTE, MUY IMPORTANTE, NUNCA y SIEMPRE parecen ser la mejor manera de dirigir al modelo lejos de minas terrestres. Espero que los modelos se vuelvan más dirigibles en el futuro y eviten esta fealdad. Pero por ahora, CC usa esto liberalmente, y tú también deberías."
393+
> — Traducido por Claude
394+
345395
### Limpiar Contexto Entre Tareas
346396

347397
Reinicia la ventana de contexto de la IA entre tareas no relacionadas para prevenir confusión y mejorar el rendimiento en nuevos problemas.

README-fr.md

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,5 @@
1-
> **Développement Actif** — Mis à jour le 5 septembre 2025
2-
> Dernière : Ajout de la technique "Posez des Questions Ouvertes, Pas Suggestives"
1+
> **Développement Actif** — Mis à jour le 11 septembre 2025
2+
> Dernière : Ajout de 6 nouvelles techniques du projet ai_stack
33
> [Voir toutes les mises à jour →](CHANGELOG.md)
44
55
**Langues disponibles :** [English](README.md) | [Español](README-es.md) | [Deutsch](README-de.md) | [Français](README-fr.md) | [日本語](README-ja.md)

README-ja.md

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,5 @@
1-
> **アクティブな開発**2025年9月5日更新
2-
> 最新:「オープンな質問をし、誘導的な質問を避ける」技術を追加
1+
> **アクティブな開発**2025年9月11日更新
2+
> 最新:ai_stackプロジェクトから6つの新しい技術を追加
33
> [すべての更新を見る →](CHANGELOG.md)
44
55
**利用可能言語:** [English](README.md) | [Español](README-es.md) | [Deutsch](README-de.md) | [Français](README-fr.md) | [日本語](README-ja.md)

README.md

Lines changed: 44 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,5 @@
1-
> **Active Development** — Updated September 5, 2025
2-
> Latest: Added "Ask Open Questions, Not Leading Ones" technique
1+
> **Active Development** — Updated September 11, 2025
2+
> Latest: Added 6 new techniques from ai_stack project
33
> [See all updates →](CHANGELOG.md)
44
>
55
> **Note:** For the best experience, visit the [website](https://coding-with-ai.dev) where you can see the popularity of each technique based on community engagement and discover which approaches developers find most valuable.
@@ -169,6 +169,13 @@ Delegate boring, systematic, and time-consuming tasks to AI - from small variabl
169169
> "The best example I've found for the agent was migrating a huge app from one UI library to another. It's not hard work, but it takes a huge amount of time and is completely uninteresting."
170170
> [Between the Prompts](https://betweentheprompts.com/design-partner/#:~:text=The%20best%20example%20I%27ve%20found%20for%20the%20agent%20was%20migrating%20a%20huge%20app%20from%20one%20UI%20library%20to%20another.%20It%27s%20not%20hard%20work%2C%20but%20it%20takes%20a%20huge%20amount%20of%20time%20and%20is%20completely%20uninteresting.)
171171
172+
### Provide Context for New Libraries
173+
174+
When using libraries outside the AI's training data, feed it recent examples and documentation to teach it how the library works.
175+
176+
> "LLMs can still help you work with libraries that exist outside their training data, but you need to put in more work—you'll need to feed them recent examples of how those libraries should be used as part of your prompt."
177+
> [Simon Willison](https://simonwillison.net/2025/Mar/11/using-llms-for-code/#:~:text=LLMs%20can%20still%20help%20you%20work%20with%20libraries)
178+
172179
### Treat AI as a Digital Intern
173180

174181
Give AI extremely precise, detailed instructions like you would to an intern - provide exact function signatures and let it handle implementation.
@@ -218,6 +225,27 @@ Have AI write comprehensive tests based on expected behavior, then iterate on im
218225
> "Ask Claude to write tests based on expected input/output pairs. Be explicit about the fact that you're doing test-driven development so that it avoids creating mock implementations, even for functionality that doesn't exist yet in the codebase. Tell Claude to run the tests and confirm they fail. Ask Claude to commit the tests when you're satisfied with them. Ask Claude to write code that passes the tests, instructing it not to modify the tests."
219226
> [Anthropic](https://www.anthropic.com/engineering/claude-code-best-practices#:~:text=Ask%20Claude%20to%20write%20tests)
220227
228+
### Ask the Agent to Review Its Own Code
229+
230+
Have the AI perform a code review on its own work before human review to surface issues and improvements.
231+
232+
> "Asking the agent to perform a code review on its own work is surprisingly fruitful."
233+
> [Chris Dzombak](https://www.dzombak.com/blog/2025/08/getting-good-results-from-claude-code/#:~:text=Asking%20the%20agent%20to%20perform%20a%20code%20review)
234+
235+
### One Writes, Another Reviews
236+
237+
Have one agent write code, then use a fresh agent to review and find problems.
238+
239+
> "Use Claude to write code. Run `/clear` or start a second Claude in another terminal. Have the second Claude review the first Claude's work. Start another Claude (or `/clear` again) to read both the code and review feedback. Have this Claude edit the code based on the feedback. This separation often yields better results than having a single Claude handle everything."
240+
> [Anthropic](https://www.anthropic.com/engineering/claude-code-best-practices#:~:text=Have%20one%20Claude%20write%20code)
241+
242+
### Edit Code in the Diff
243+
244+
Review changes in diff view and type corrections directly into the diff before committing.
245+
246+
> "I manually review all AI-written code and test cases. I'll add test cases for anything I think is missing or needs improvement, either manually or by asking the LLM to write those cases (which I then review)."
247+
> [Chris Dzombak](https://www.dzombak.com/blog/2025/08/getting-good-results-from-claude-code/#:~:text=I%20manually%20review%20all)
248+
221249
## Cross-Stage Techniques
222250

223251
### Run Multiple Agents in Parallel
@@ -251,13 +279,27 @@ Create tools that respond quickly, provide clear error messages, and protect aga
251279
> "Tools need to be fast. The quicker they respond (and the less useless output they produce) the better. Crashes are tolerable; hangs are problematic. Tools need to be user friendly! Tools must clearly inform agents of misuse or errors to ensure forward progress. Tools need to be protected against an LLM chaos monkey using them completely wrong. There is no such thing as user error or undefined behavior!"
252280
> [Armin Ronacher](https://lucumr.pocoo.org/2025/6/12/agentic-coding/#:~:text=Tools%20need%20to%20be%20fast)
253281
282+
### Start Cheap, Escalate When Stuck
283+
284+
Begin with faster/cheaper models for routine tasks, then escalate to more powerful models only when you hit complex problems.
285+
286+
> "Sonnet 4 handles 90% of tasks effectively. Switch to Opus when Sonnet gets stuck. Recommend starting with Sonnet and providing comprehensive context."
287+
> [Sankalp](https://sankalp.bearblog.dev/my-claude-code-experience-after-2-weeks-of-usage/#:~:text=Sonnet%204%20handles%2090%25)
288+
254289
### Clear Context Between Tasks
255290

256291
Reset the AI's context window between unrelated tasks to prevent confusion and improve performance on new problems.
257292

258293
> "During long sessions, Claude's context window can fill with irrelevant conversation, file contents, and commands. This can reduce performance and sometimes distract Claude. Use the `/clear` command frequently between tasks to reset the context window."
259294
> [Anthropic](https://www.anthropic.com/engineering/claude-code-best-practices#:~:text=During%20long%20sessions%2C%20Claude's%20context)
260295
296+
### Use Strong Emphasis in Prompts
297+
298+
Use IMPORTANT, NEVER, ALWAYS liberally in prompts to steer AI away from common mistakes - it's still the most effective approach.
299+
300+
> "Unfortunately CC is no better when it comes to asking the model to not do something. IMPORTANT, VERY IMPORTANT, NEVER and ALWAYS seem to be the best way to steer the model away from landmines. I expect the models to get more steerable in the future and avoid this ugliness. But for now, CC uses this liberally, and so should you."
301+
> [Vivek (MinusX AI Team)](https://minusx.ai/blog/decoding-claude-code/#:~:text=Unfortunately%20CC%20is%20no%20better)
302+
261303
### Interrupt and Redirect Often
262304

263305
Don't let AI go too far down the wrong path - interrupt, provide feedback, and redirect as soon as you notice issues.

0 commit comments

Comments
 (0)