Noah's dad 😻 Nature lover 🌷 Software engineer 👨🏻💻 Pilot 👨🏻✈️ In a timeless discovery journey called life 🏞️
Somewhere in my mind
Joined July 2008
- Tweets7.3K
- Following3.3K
- Followers259
- Likes40.5K
Pinned Tweet
É a maneira como você encara a vida que faz toda a diferença! #reflexão
Alessandro Gambin da Silva retweeted
How Garbage Collectors Actually Work
Automatic memory management is not magic. It is a set of carefully engineered tradeoffs between throughput, latency, and memory overhead.
Alessandro Gambin da Silva retweeted
A context switch is smaller than it sounds:
1. push the callee-saved registers
2. save the stack pointer into the old thread
3. load the new thread's stack pointer
4. pop its registers
5. ret
The ret returns into a different thread.
Alessandro Gambin da Silva retweeted
Hands-On Docker Roadmap 🐳
- Over 100 practical problems, from simple to advanced
- Realistic scenarios with automatic checks, hints, and editorial solutions
- 0% chance of getting into a tutorial hell trap
- Extra focus on the underlying Linux topics
labs.iximiuz.com/roadmaps/do…
Alessandro Gambin da Silva retweeted
𝗠𝗖𝗣 𝘃𝘀 𝗔𝗣𝗜.
An 𝗔𝗣𝗜 defines how software systems communicate through specific endpoints, requests, and responses. It gives applications a structured way to access data or trigger functionality in another system.
𝗠𝗖𝗣 gives AI applications a standardized way to discover and use external tools, data, and resources. Instead of building custom integrations for every AI client, an MCP server exposes capabilities through a common protocol.
APIs expose functionality to software. MCP standardizes how AI applications discover and interact with that functionality.
But once AI sits behind an API, the request-response model gets harder.
Inference might take longer than the request can stay open. It might fail halfway through. It might need to be retried.
That changes how the API itself should be designed.
Oracle’s guide breaks down how to design for that with asynchronous jobs, workers, durable state, and predictable API contracts.
𝗥𝗲𝗮𝗱 𝘁𝗵𝗲 𝗴𝘂𝗶𝗱𝗲 → lucode.co/rest-api-for-ai-ap…
What else would you add?
——
🙏 Thanks to @OracleDevs for sponsoring this post.
➕ Follow me ( Nikki Siapno ) to improve at AI and system design.
𝗠𝗖𝗣 𝘃𝘀 𝗔𝗣𝗜.
An 𝗔𝗣𝗜 defines how software systems communicate through specific endpoints, requests, and responses. It gives applications a structured way to access data or trigger functionality in another system.
𝗠𝗖𝗣 gives AI applications a standardized way to discover and use external tools, data, and resources. Instead of building custom integrations for every AI client, an MCP server exposes capabilities through a common protocol.
APIs expose functionality to software. MCP standardizes how AI applications discover and interact with that functionality.
But once AI sits behind an API, the request-response model gets harder.
Inference might take longer than the request can stay open. It might fail halfway through. It might need to be retried.
That changes how the API itself should be designed.
Oracle’s guide breaks down how to design for that with asynchronous jobs, workers, durable state, and predictable API contracts.
𝗥𝗲𝗮𝗱 𝘁𝗵𝗲 𝗴𝘂𝗶𝗱𝗲 → lucode.co/rest-api-for-ai-ap…
What else would you add?
——
🙏 Thanks to @OracleDevs for sponsoring this post.
➕ Follow me ( Nikki Siapno ) to improve at AI and system design.
Alessandro Gambin da Silva retweeted
Most devs have never built:
• a tokenizer
• a memory allocator
• a scheduler
• a syscall handler
LowLevelCraft has 479+ micro-tasks that make you build all of them.
In the browser. No setup. C, C++, Rust, Python, ARM asm.
The C track is free 👇
Alessandro Gambin da Silva retweeted
We're adding a new plugin to Claude Code: You should Know.
It scans Claude's output for important information you might miss to help keep you in the loop.
Enable it with:
/plugin enable cc-plugin-you-should-know@builtin
Alessandro Gambin da Silva retweeted
On public demand, I will upload the full guide tomorrow, so please share this and follow for more content like this.
Alessandro Gambin da Silva retweeted
We have a Stitch MCP. We have a Stitch SDK. But, we don't have a Stitch CLI. Well, not until now.
Introducing the @google/stitch CLI:
🔷 Connect to your local coding agents
🔷 Generate screens and design systems
🔷 Send a local dev server snapshot to Stitch
Do it all without leaving the terminal or better yet, ask your favorite harness like @antigravity 😎
Learn more 👇
Alessandro Gambin da Silva retweeted
What is RBAC (Role-Based Access Control)?
It's one of the most flexible auth policies you can implement.
Here's how it can help with your authorization requirements.
RBAC stands for Role-Based Access Control:
- Roles define high-level policies
- Roles have a set of permissions
- Users belong to roles and inherit permissions
- Permissions decide what a user can or can't do
With standard role-based authorization, I always had a problem with fine-grained auth rules.
Roles are too broad for some policies you'll typically want to enforce.
This is the problem permissions solve.
I can define permissions for specific actions.
It's easy to allow other roles access: you assign the permission to the role.
If you want to learn more about RBAC authorization and how to implement it, start here: milanjovanovic.tech/blog/bui…
RBAC makes auth into a flexible, scalable system. Start with permissions. Define what actions users can perform, not what roles they have. It's a much easier way to reason about your app's actions.
Alessandro Gambin da Silva retweeted
8 habits in Claude Code burning your tokens and your code quality right now
And how to fix them
1. One big task, one session.
Claude re-reads the entire context on every single message.
↳ The session you kept alive for 3 days is slower, pricier, and worse at following your own conventions.
2. Claude doesn't read your whole repo.
At session start it knows two things: your CLAUDE[.]md and the file tree. Nothing else.
↳ Type @ and pin the exact file you mean. Every wrong guess costs you tokens and quality. As Claude will search more files to find what it needs.
3. Match the model to the job.
Haiku for a rename. Sonnet for one small feature. Opus for a list of features, complex feature, or fixing a complex problem. Fable for creating a plan, making code review, and for the hardest problems.
↳ Max effort on a simple task doesn't buy quality. Claude can write overcomplicated code instead.
4. Write down the patterns you don't use.
Claude has seen a million repos with AutoMapper, MediatR and the Repository pattern.
↳ If "Patterns We Do NOT Use" is missing from your CLAUDE[.]md, they land in your code immediately.
5. Keep CLAUDE[.]md under 100 lines.
Every line is loaded into every message of every session, relevant or not.
↳ A bloated rulebook doesn't make Claude smarter. It buries the rules that matter. Move other details on implementing features to Claude skills.
6. The plan should live in the file
A markdown file in a plans/ folder survives the session, the laptop reboot, and can be reused by your team. A plan saved in the session's memory can be lost.
↳ Copy it into the frontend repo, and a fresh session continues like it was there all along.
7. Pre-approve the boring commands.
dotnet build, dotnet test and git diff belong in the allow list. git push and appsettings.Production.json belong in deny.
↳ If you need to approve 20 prompts in each session - you train yourself to click yes without reading.
8. Delegate the research.
"How does authorization flow through this API" can mean reading 30 files.
↳ Send a subagent to read them in its own window and hand back the answer. Your context stays clean.
The gap between fighting Claude Code and shipping with it is one weekend of setup.
👉 P.S. I packaged the full system into a free 10-lesson email course: Claude Code for .NET Developers.
Plan files, CLAUDE[.]md templates, skills, hooks, subagents, and PR reviews running in CI.
Grab it here: antondevtips.com/claude-code…
Which one are you fixing first: the session that never ends, or big CLAUDE[.]md?
Alessandro Gambin da Silva retweeted
a maioria dos devs começam diversos projetos de gaveta com IDEAS incríveis que ninguém quer pagar
essa já foi minha história, mas o que sucesso é quase como uma receita de bolo
você não parte do produto, você não parte da solução…. o ponto de partida é uma DOR
deixa eu exemplificar usando a enxovaly
DOR: mãs de primeira viagem não sabem o que comprar para o bebê
PRODUTO: um app
SOLUÇÃO: lista de enxoval personalizada
PROMESSA: ajudar mães de primeira viagem a comprar o enxoval do bebê sem exageros
perceba que o app é só o produto, poderia ser uma planilha de excel, poderia ser um ebook/pdf, isso não importa
porque o que as pessoas tão comprando é a promessa, aquilo que vai resolver a DOR
a questão é que a GRANDE MAIORIA, dos Saas e apps falham, porque não resolvem dor nenhum, são só ideias legais
entenda esse conceito, isso é o ponto de partida que a MAIORIA pula, porque é chato, e DESISTE, porque fez um app INUTIL que ninguem quer pagar
Alessandro Gambin da Silva retweeted
Coisas que tornam um Dev produtivo de verdade
- saber usar o terminal
- saber debugar
- ler código rápido
- escrever pouco e claro
- saber pedir ajuda cedo
o que mais?
Alessandro Gambin da Silva retweeted
Coisas que Dev só aprende com incidente em produção
- ter rollback
- ter alerta que importa
- ter log com contexto
- ter runbook
- não subir sem feature flag
o que mais?
Alessandro Gambin da Silva retweeted
It's over
Aqui na empresa utilizam o Claude Enterprise, que a partir de Outubro será cobrado via pricing de API
Definiram algumas faixas de limite, a do meu time é de $900 em tokens/mês
O lado bom é que eu só tenho utilizado Opus low/medium pra quase tudo hoje em dia - aprender a ser token efficient será importante para quando todas as empresas chegarem na época de vacas magras
Nos nossos projetos pessoais esquecemos que o allowance é bem mais alto, e portanto pensamos que qualquer um consegue chegar em certo resultado
Isso não é verdade quando o limite é outro e quando você precisa fazer aquele investimento valer
Sinto que depender de frontier model em high/xhigh/max reasoning é skill issue - qualquer tarefa pode ser quebrada o suficiente para ser implementada por um modelo bem menor
Em partes, a limitação irá nos fazer aprender como ser eficientes, mas também podemos exigir mais budget a depender de um processo de aprovação
Aproveitem enquanto é tempo, a época de tokens magros está chegando
Alessandro Gambin da Silva retweeted
50 legendary internet rabbit holes you can spend hours exploring👇
1. zoom.earth — Watch the world via live satellite imagery
2. flightradar24.com — See every plane currently in the sky
3. marinetraffic.com — Track all ships at sea in real time
4. windy.com — Live map of winds and storms
5. lightningmaps.org — Watch lightning strikes hitting Earth in real time
6. earthquake.usgs.gov — Live list of recent earthquakes
7. submarinecablemap.com — Ocean cables carrying the internet
8. globalforestwatch.org — Watch forests disappear from space
9. worldometers.info — The world's statistics, second by second
10. internetlivestats.com — Current number of tweets and searches being posted
11. thetruesize.com — Compare the true sizes of countries
12. oldmapsonline.org — Maps from centuries ago
13. davidrumsey.com — Archive of 150,000 historical maps
14. openstreetmap.org — World map drawn by volunteers
15. window-swap.com — Look out the window of a random person around the world
16. virtualvacation.us — Virtual walks through cities
17. mapcrunch.com — Teleport to a random spot on Earth
18. atlasobscura.com — Catalog of the world's strangest places
19. neal.fun — Interactive knowledge experiences
20. htwins.net/scale2 — Scale journey from atom to universe (Updated HTML5 link)
21. eyes.nasa.gov — Explore the solar system in 3D
22. stellarium-web.org — Real sky map in your browser
23. apod.nasa.gov — NASA's astronomy picture of the day
24. images.nasa.gov — NASA's entire visual archive, free
25. pudding.cool — Visual articles told through data
26. ourworldindata.org — The state of the world with real data
27. gapminder.org — What we mistakenly think we know about the world
28. informationisbeautiful.net — Visualizing complex data
29. data.worldbank.org — World Bank's open data
30. data.tuik.gov.tr — Turkey's official statistics database
31. archive.org — Archive of millions of books, films, and software
32. gutenberg.org — 70,000 free books whose copyrights have expired
33. openlibrary.org — Record of every book in the world
34. loc.gov — U.S. Library of Congress digital archive
35. europeana.eu — Europe's cultural heritage archive
36. dp.la — America's digital library collection
37. artsandculture.google.com — Tour museums from home
38. rijksmuseum.nl — Download artworks in high resolution
39. wikiart.org — Archive of 250,000 artworks
40. publicdomainreview.org — Forgotten visual treasures of history
41. openculture.com — Free archive of culture and education
42. metmuseum.org — Met Museum's open collection
43. musicmap.info — Family tree of music genres
44. radiooooo.com — Pick a country and decade to listen to that era
45. listen.hatnote.com — Turn Wikipedia edits into audio
46. wikipedia.org — Random knowledge well
47. timeanddate.com — Time, sunrises, and sky events
48. sciencedaily.com — Live stream of science news
49. arxiv.org — Free preprints of scientific papers
50. observablehq.com — Visualize data with live code
Save this. You’ll definitely need some of these later. 🔖
Follow @justinbrave21 for more useful websites, AI tools & tech resources.
Alessandro Gambin da Silva retweeted
Netflix replaced their 15 years old recommendation algorithm with an LLM.
It’s called "GenRec" and it completely changes how recommendation algorithms are built.
For over a decade, the Netflix recommendation engine was a masterclass in feature engineering. Data scientists built thousands of complex, handcrafted features to figure out what you wanted to watch next.
It required bespoke architectures. Massive infrastructure. Constant manual tuning.
Netflix threw all of it away.
They built GenRec, an LLM-backed ranker.
Instead of translating your behavior into complex math, they just turn your watch history and metadata into a natural language sentence.
They feed that raw text into a foundation LLM.
The AI simply reads your behavior like a story, understands your evolving tastes, and scores the entire catalog in a single forward pass.
No manual feature engineering. No complex bespoke architectures.
Here is the part that should terrify traditional data scientists.
This text-based LLM didn't just match the highly tuned production system Netflix spent years perfecting.
It beat it.
And it achieved those statistically significant gains using roughly 40x fewer labeled training examples.
We are watching a massive paradigm shift in real time.
The most complex predictive algorithms in the world are being replaced by models that just know how to read.
If Netflix can replace their core product engine with an LLM, what complex system in your business is about to become obsolete?
Alessandro Gambin da Silva retweeted
𝗧𝗿𝗮𝗱𝗶𝘁𝗶𝗼𝗻𝗮𝗹 𝗖𝗼𝗱𝗲 𝗥𝗲𝘃𝗶𝗲𝘄 𝘃𝘀 𝗔𝗴𝗲𝗻𝘁𝗶𝗰 𝗖𝗼𝗱𝗲 𝗥𝗲𝘃𝗶𝗲𝘄.
As coding agents generate more code, the bottleneck is shifting from writing it to reviewing it.
In a 𝘁𝗿𝗮𝗱𝗶𝘁𝗶𝗼𝗻𝗮𝗹 𝘄𝗼𝗿𝗸𝗳𝗹𝗼𝘄, a developer writes code, opens a PR, automated checks run, and a human reviews the changes. Feedback goes back to the developer until the PR is ready to merge.
𝗔𝗴𝗲𝗻𝘁𝗶𝗰 𝗰𝗼𝗱𝗲 𝗿𝗲𝘃𝗶𝗲𝘄 moves that loop earlier and makes more of it autonomous.
A coding agent generates a change → a separate review agent inspects it → findings feed back into another revision → the change is reviewed again.
That can start before a PR is even opened, then continue through CI and the PR workflow.
But reviewing each change is only part of the problem.
As change volume grows, 𝘁𝗲𝗮𝗺𝘀 𝗮𝗹𝘀𝗼 𝗻𝗲𝗲𝗱 𝘁𝗼 𝗱𝗲𝗰𝗶𝗱𝗲 what deserves attention, understand what actually changed, and manage the risk of what ships.
That's the broader idea behind 𝗮𝗴𝗲𝗻𝘁𝗶𝗰 𝗰𝗵𝗮𝗻𝗴𝗲 𝗺𝗮𝗻𝗮𝗴𝗲𝗺𝗲𝗻𝘁.
CodeRabbit extends beyond AI code review with 𝗧𝗿𝗶𝗮𝗴𝗲 to prioritize changes, 𝗖𝗵𝗮𝗻𝗴𝗲 𝗦𝘁𝗮𝗰𝗸 to explain complex changes, and 𝗦𝗲𝗰𝘂𝗿𝗶𝘁𝘆 to find and verify risks across the codebase.
The real shift is from reviewing code line by line to reviewing change as a system.
𝗦𝘁𝗮𝗿𝘁 𝗮 𝗳𝗿𝗲𝗲 𝟭𝟰-𝗱𝗮𝘆 𝘁𝗿𝗶𝗮𝗹, 𝗻𝗼 𝗰𝗿𝗲𝗱𝗶𝘁 𝗰𝗮𝗿𝗱, 𝟮-𝗰𝗹𝗶𝗰𝗸 𝘀𝗲𝘁𝘂𝗽 → lucode.co/coderabbit-agentic…
What else would you add?
——
♻️ Repost to help others learn AI.
🙏 Thanks to @coderabbitai for sponsoring this post.
➕ Follow me ( Nikki Siapno ) to improve at AI and system design.
Alessandro Gambin da Silva retweeted
These are the best visual AI resources for learning Transformers, LLMs, embeddings, diffusion, inference and model internals ↓
1/ Transformer Explainer - Watch GPT process text through embeddings, attention, MLPs and next-token prediction.
poloclub.github.io/transform…
2/ Brendan Bycroft’s LLM Visualization - Explore an LLM from architecture down to tensors and operations.
bbycroft.net/llm
3/ 3Blue1Brown - Visual intuition for linear algebra, neural networks, backprop, attention and Transformers.
3blue1brown.com/topics/neura…
4/ The Illustrated Transformer - One of the clearest visual explanations of embeddings, Q/K/V and attention.
jalammar.github.io/illustrat…
5/ TensorFlow Playground - Watch neural networks learn as you change layers, activations, features and learning rate.
playground.tensorflow.org/
6/ Google PAIR AI Explorables - Interactive explainers on LLMs, generalization, interpretability and model behavior.
playground.tensorflow.org/
7/ Distill - Exceptional visual essays on t-SNE, feature visualization, GNNs and interpretability.
distill.pub/
8/ Abhik Sarkar’s Transformer Visualizations - RoPE, KV cache, FlashAttention, MQA, GQA and more.
playground.tensorflow.org/
9/ Visual Guide to Attention Variants - MHA, MQA, GQA and MLA visually compared.
magazine.sebastianraschka.co…
10/ Visual Guide to Mixture of Experts - Routing, experts, sparse activation and load balancing.
newsletter.maartengrootendor…
11/ Modular LLM Inference Handbook - Prefill, decode, KV cache, batching, quantization and speculative decoding.
handbook.modular.com/llm-inf…
12/ Apple Embedding Atlas - Explore clusters, neighborhoods and outliers in large embedding spaces.
apple.github.io/embedding-at…
13/ Diffusion Explainer - Follow Stable Diffusion step by step.
poloclub.github.io/diffusion…
14/ Neuronpedia - Explore features, activations, SAE latents and attribution graphs inside real models.
neuronpedia.org/
15/ Seeing Theory - Probability, Bayes, distributions, regression and inference made interactive.
seeing-theory.brown.edu/
16/ CNN Explainer - Visualize convolutions, feature maps, activations and pooling.
poloclub.github.io/cnn-expla…
17/ GAN Lab - Train a GAN in your browser and watch its generated distribution evolve.
poloclub.github.io/ganlab/
Save this. There’s a serious AI curriculum hiding inside these links.