Soumya Nayak

I do fluent Sarcasm!!!😜

[Avi Chawla @_avichawla]

The anatomy of a perfect /goal prompt in Claude Code!

(save for later and use it to reduce token usage)

The way /goal works is that two models run in a loop.

- Sonnet/Opus handles the actual coding.
- After every turn, Haiku reads the full conversation transcript and checks whether completion conditions were met.
- If Haiku says "not yet," Sonnet/Opus starts another turn automatically.

The Haiku evaluator costs almost nothing. But the worker still runs at full Sonnet/Opus rates.

So the condition you write directly controls how many expensive turns Claude takes before the loop terminates.

Anthropic recommends three things in a /goal condition:
- a measurable end state
- a stated check
- and constraints.

They also suggest adding a turn cap.

What the docs don't cover is what happens when any of those pieces are missing.

A vague condition like "make the app production-ready" loops indefinitely because no command output ever proves that claim.

A subjective condition like "complete the migration" lets Claude self-report completion, and Haiku accepts it because the judge only sees the transcript.

The nine-section template in the image expands Anthropic's three-part recommendation into a structured prompt that addresses both failure modes.

→ GOAL defines the objective in a single sentence, eliminating ambiguity.

→ CONTEXT gives Claude the background it needs, so it doesn't waste turns exploring the codebase.

→ CONSTRAINTS set hard scope boundaries.

→ PRIORITY controls execution order so Claude tackles easy wins first.

→ PLAN states the approach explicitly, preventing aimless exploration that burns tokens.

→ DONE WHEN forces a binary, observable outcome like pytest exits 0, not a subjective state like "code quality is improved."

→ VERIFY makes Claude run a specific command whose raw output lands in the transcript, so Haiku judges machine evidence, not self-reports.

→ OUTPUT specifies what Claude should surface when done.

→ STOP RULES cap the downside with a turn limit. Without this, a slightly wrong condition loops forever.

For even stronger guarantees, pair /goal with a Stop hook that runs your test suite or hits a CI endpoint after every turn. That moves verification from the prompt layer to the infrastructure layer, where it's deterministic.

I found this template on findskill(.)ai while researching the perfect way to use /goal command. They don't claim to have created it either, but mentioned it as a community-created template that circulated on X.
____
Find me → @_avichawla
Every day, I share tutorials and insights on DS, ML, LLMs, and RAGs.

[Miguel Ángel Durán @midudev]

Este repo es una joya: tutoriales crear agentes de IA.
Listos para producción y con casos de uso reales.
Todo el código disponible y explica cómo desplegarlos:
https://t.co/Bf4WLIdNmM

[Dhaval Makwana @heyDhavall]

List of 9 websites where you can download unlimited books for free:

1. smashwords .com
2. baen .com
3. feedbooks .com/publicdomain
4. wikibooks .org
5. google books
6. openstax .org
7. bookboon .com
8. freecomputerbooks .com
9. standardebooks .org

[Ashish Pratap Singh @ashishps_1]

30 Blogs to Learn 30 System Design Concepts:

1. Scalability: https://t.co/NOD5gy8m5G

2. Availability: https://t.co/uiTwjsP7M3

3. Consistent Hashing: https://t.co/MWhzdMY1c0

4. CAP Theorem: https://t.co/yviUBJLkWG

5. TCP vs UDP: https://t.co/j5DCt7o6IY

6. DNS: https://t.co/hXTf9L41B8

7. Proxy vs Reverse Proxy: https://t.co/TWzR2nL2CC

8. Load Balancing Algorithms: https://t.co/zrVmh7ssGI

9. APIs: https://t.co/bmxBOB6l6F

10. API Gateways: https://t.co/xnEU5ZNqhL

11. Rate Limiting: https://t.co/mjnf3SanK8

12. JWT: https://t.co/IPT4Z47GJA

13. WebSockets: https://t.co/MmKfGEpYLa

14. Webhooks: https://t.co/WjRDQAXiNp

15. Message Queues: https://t.co/pkivlCr36P

16. Change Data Capture (CDC): https://t.co/uh7BoHeUNw

17. Caching Strategies: https://t.co/ogkqglKQwI

18. Content Delivery Network (CDN): https://t.co/WO50R2FRIt

19. Database Types: https://t.co/yOFTzDa7BE

20. ACID Transactions: https://t.co/TpWnHfvS3Y

21. Indexing: https://t.co/RrHGXADOLK

22. Sharding: https://t.co/VxbGTbDNRm

23. Block vs File vs Object Storage: https://t.co/KJIt0JIu7f

24. Strong vs Eventual Consistency: https://t.co/MkR9wU8LZI

25. Service Discovery: https://t.co/UurGannLZD

26. Batch vs Stream Processing: https://t.co/qYT3JbmFrx

27. Observability: https://t.co/zyiteDQNJz

28. Authentication & Authorization: https://t.co/oo9YxuomNr

29. Idempotency: https://t.co/psi0hY88bC

30. GeoHashing: https://t.co/iSTGo48IYa

♻️ Repost to help others in your network.

[Neo Kim @systemdesignone]

If you want to boost your software engineering career, star + bookmark these 12 GitHub repos:

1 Vision Agent
↳ https://t.co/M7OIzZVAKA

2 System design
↳ https://t.co/57FAu0f9Oz

3 Public APIs
↳ https://t.co/Ikt9Zyy2m4

4 Tech interview handbook
↳ https://t.co/3ZpaTKgDzE

5 Coding interview university
↳ https://t.co/cQhGu0zI2E

6 Engineering leadership
↳ https://t.co/i7vTHC5Foz

7 Freecodecamp
↳ https://t.co/HbP9utSuZE

8 Developer roadmaps
↳ https://t.co/RXQVuYKr8z

9 Path to senior engineer handbook
↳ https://t.co/X4kepQW20p

10 Free programming books
↳ https://t.co/WuauWJKWQ6

11 Build your own x
↳ https://t.co/gW3qDjnCxa

12 Awesome lists
↳ https://t.co/fdOmRT2iLF

Remember, star + bookmark these repos right now.

(What else would you add?)

——

💾 Save this for later & RT to help others become good software engineers.

👤 Follow @systemdesignone + turn on notifications.

[Neo Kim @systemdesignone]

If you want to get started with SYSTEM DESIGN (in 2026), learn these 20 concepts:

1 How Idempotent API Works
↳ https://t.co/afe7ACuSYE

2 Saga Design Pattern
↳ https://t.co/2CffTodOHL

3 Redis Use Cases
↳ https://t.co/hZ571ruVeA

4 How JWT Works
↳ https://t.co/SZXXrlBsWH

5 How RPC Actually Works
↳ https://t.co/yeIgcmAxQx

6 How Consistent Hashing Works
↳ https://t.co/7d6EipPcKF

7 How Service Discovery Works
↳ https://t.co/BcL3tgxx1u

8 Monolith vs Microservices Architecture
↳ https://t.co/KwVAEGVkA9

9 What Happens When You Type a URL Into Your Browser
↳ https://t.co/P3SiURMFlW

10 Microservices Lessons From Netflix
↳ https://t.co/XgS7VQoBFv

11 Frontend System Design 101
↳ https://t.co/ViPOQrLZzA

12 How Websockets Work
↳ https://t.co/JfT6mj4mrv

13 How Bloom Filters Work
↳ https://t.co/ntZXq7LxVn

14 How Databases Keep Passwords Securely
↳ https://t.co/KSfIhpAT2j

15 How Does HTTPS Work
↳ https://t.co/r5rUtVpw0O

16 Modular Monolith Architecture
↳ https://t.co/VVV6v3KGHJ

17 API Design Best Practices
↳ https://t.co/I2ejJ0kbYq

18 How DNS Works
↳ https://t.co/H7hcZnws8N

19 System Design 101
↳ https://t.co/DgL8xz0KTQ

20 API Versioning - A Deep Dive
↳ https://t.co/OHAtKSUgVN

(What else should make this list?)

——

👋 PS - Want my System Design Playbook for FREE?

Join my newsletter with 200K+ software engineers:

→ https://t.co/ByOFTtOihX

———

💾 Save this for later, and RT to help others learn system design.

👤 Follow @systemdesignone + turn on notifications.

[Tom Dörr @tom_doerr]

Implementations of 17+ agentic architectures

https://t.co/l03dJU2k8p

[Akshay 🚀 @akshay_pachaar]

LLM fine-tuning techniques I'd learn if I were to customize them:

Bookmark this.

1. LoRA
2. QLoRA
3. Prefix Tuning
4. Adapter Tuning
5. Instruction Tuning
6. P-Tuning
7. BitFit
8. Soft Prompts
9. RLHF
10. RLAIF
11. DPO (Direct Preference Optimization)
12. GRPO (Group Relative Policy Optimization)
13. RLAIF (RL with AI Feedback)
14. Multi-Task Fine-Tuning
15. Federated Fine-Tuning

My favourite is GRPO for building reasoning models. What about you?

I've shared my full tutorial on GRPO in the replies.

[Soumya Nayak @isudo]

@boringdev17 Send and repost

[Cicero @cicerougc]

A senior Google engineer just dropped a 400-page free book on Agentic Design Patterns (Google Docs).

Table of contents looks wild:
• Advanced Prompt Techniques
• Multi-Agent Patterns
• Everything you need to know about agents + code

Reply send and repost, and I will dm it.

[Avi Chawla @_avichawla]

Deploy AI apps by just adding a Python decorator (open-source)!

Beam is an open-source alternative to Modal for deploying serverless AI workloads.

Simply uv add beam-client, add a decorator, and turn any workflow into a serverless endpoint!

[Unwind AI @unwind_ai_]

RAG is not Memory for AI Agents.

5 AI memory engines to build agents that maintain long-term context and learn continuously:

(Last 2 released just this month)

1. Zep builds and queries temporally-aware knowledge graphs that evolve with every interaction.

100% Opensource.

[Avi Chawla @_avichawla]

voyage-context-3 embedding model by @MongoDB solves this.

It is a contextualized chunk embedding model that produces vectors for chunks that capture the full document context without any manual metadata and context augmentation.

Check this visual 👇

[ClickHouse @ClickHouseDB]

Talk to your database like it’s ChatGPT.

The new MCP Server makes ClickHouse work seamlessly with LLMs.

Details @ https://t.co/7CfuN7dY0N

[Louis Gleeson @aigleeson]

I finally understand how large language models actually work

After reading the 2025 textbook “Foundations of LLMs”

It blew my mind and cleared up years of confusion

Here’s everything i learned (in plain english):

[Akshay 🚀 @akshay_pachaar]

Make Claude Code 10x more powerful!

Code Context is an MCP plugin that brings semantic code search to Claude Code, Gemini CLI, or any AI coding agent.

Full codebase indexing means richer context and better code generation.

100% open-source.

[Ashish Pratap Singh @ashishps_1]

30 Blogs to Learn 30 System Design Concepts:

1. APIs: https://t.co/SFR7c1yi1Z

2. JWTs: https://t.co/hHSdPVqsML

3. Webhooks: https://t.co/789CPy1rXE

4. API Gateways: https://t.co/kjlIJy9ElI

5. Load Balancing: https://t.co/zrVmh7rURa

6. Proxy vs Reverse Proxy: https://t.co/AR8L9F2met

7. Scalability: https://t.co/Gbpjax24gd

8. Availability: https://t.co/GfakejdGQz

9. SPOF: https://t.co/5FtrH1Shrc

10. CAP Theorem: https://t.co/Aaq5s96WFU

11. SQL vs NoSQL: https://t.co/vDTE5PqNel

12. ACID Transactions: https://t.co/eQBBIIM6w9

13. Database Indexes: https://t.co/CKk39deBkk

14. Database Sharding: https://t.co/nIkBZ8FZtv

15. Consistent Hashing: https://t.co/cg1mZ4iuQy

16. CDC: https://t.co/YCesZ6Y8AN

17. Caching: https://t.co/Jed9uIFFim

18. Caching Strategies: https://t.co/xhxmMgVQvW

19. Cache Eviction Policies: https://t.co/eGUV3zzG3A

20. CDN: https://t.co/DVnDPGUsFC

21. Rate Limiting Algorithms: https://t.co/mjnf3S9PUA

22. Message Queues: https://t.co/BTX85efXZu

23. Bloom Filters: https://t.co/vNV9yFrixf

24. Idempotency: https://t.co/lsMSkZrS38

25. Concurrency vs Parallelism: https://t.co/DWmMDNjkxR

26. Long Polling vs WebSockets: https://t.co/MB628lZb2w

27. Stateful vs. Stateless Architecture: https://t.co/w7OkmUQLR8

28. Batch vs Stream Processing: https://t.co/q3MCCs26wd

29. Strong vs. Eventual Consistency: https://t.co/acdTNzakRK

30. REST vs GraphQL: https://t.co/p0aR5Xxnk3

♻️ Repost to help others in your network.

[Julian Goldie SEO @JulianGoldieSEO]

N8N FULL COURSE 5 HOURS (Build & Automate Anything)

Want the full guide? DM me.

[Aurimas Griciūnas @Aurimas_Gr]

Building a 𝗽𝗿𝗼𝗱𝘂𝗰𝘁𝗶𝗼𝗻 𝗴𝗿𝗮𝗱𝗲 𝗥𝗲𝘁𝗿𝗶𝗲𝘃𝗮𝗹 𝗔𝘂𝗴𝗺𝗲𝗻𝘁𝗲𝗱 𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝗼𝗻 (𝗥𝗔𝗚) 𝗯𝗮𝘀𝗲𝗱 𝗔𝗜 𝘀𝘆𝘀𝘁𝗲𝗺 is a difficult task. Don't miss the 𝘏) and read until the end to understand why 👇

Here are some of the moving parts in the RAG based systems that you will need to take care of and continuously tune in order to achieve desired results:

𝗥𝗲𝘁𝗿𝗶𝗲𝘃𝗮𝗹:

𝘍 ) Chunking - how do you chunk the data that you will use for external context.

- Small, Large chunks.
- Sliding or tumbling window for chunking.
- Retrieve parent or linked chunks when searching or just use originally retrieved data.

𝘊 ) Choosing the embedding model to embed and query and external context to/from the latent space. Considering Contextual embeddings.

𝘋 ) Vector Database.

- Which Database to choose.
- Where to host.
- What metadata to store together with embeddings.
- Indexing strategy.

𝘌 ) Vector Search

- Choice of similarity measure.
- Choosing the query path - metadata first vs. ANN first.
- Hybrid search.

𝘎 ) Heuristics - business rules applied to your retrieval procedure.

- Time importance.
- Reranking.
- Duplicate context (diversity ranking).
- Source retrieval.
- Conditional document preprocessing.

𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝗼𝗻:

𝘈 ) LLM - Choosing the right Large Language Model to power your application.

✅ It is becoming less of a headache the further we are into the LLM craze. The performance of available LLMs are converging, both open source and proprietary. The main choice nowadays is around using a proprietary model or self-hosting.

𝘉 ) Prompt Engineering - having context available for usage in your prompts does not free you from the hard work of engineering the prompts. You will still need to align the system to produce outputs that you desire and prevent jailbreak scenarios.

And let’s not forget the less popular part:

𝘏) Observing, Evaluating, Monitoring and Securing your application in production!

What other pieces of the system am I missing? Let me know in the comments 👇

#LLM #AI #MachineLearning

[Soumya Nayak @isudo]

Boost your testing game with GitHub Copilot! Learn tips and see examples on how to generate unit tests efficiently using AI. #GitHubCopilot #UnitTesting #AI

https://t.co/JzKX7tJ4UA