DEJ retweeted
AI agent guardrails matter a lot!
There’s 3 types of guardrails:
- input guardrails
We need to validate the user input and if they say something inappropriate or off topic, reject it without sending it to the agent
- output guardrails
Sometimes for a perfectly appropriate input, the agent generates an inappropriate output. Validate this output before sending it to the user and tell them when an inappropriate output was generated
- tool guardrails
Tools need protection and limits. Rate limit guardrails are the most important one!
Follow me for more AI tips and tricks!
DEJ retweeted
As a continuation to this project, I'm currently working on something fun on top of it.
Introducing moochy chat: a local AI chat that runs entirely in your browser tab(CPU or GPU), meaning you don't need internet connection to use it after model gets downloaded in the browser tab.
got4 is < 54Kb of Web Assembly.
The page first load is 135Kb.
For some models you can go up to 150 tokens per seconds.
No subscription, No API key, 100% free.
For now, faster than WebLLM(@mlc_ai_ ), wllama(@ngxson ) and Transformers.js(@xenovacom) on benchmarks.
Models shipped right now are : @OpenBMB (MiniCPM), @Alibaba_Qwen, @MistralAI, @AIatMeta (Llama), @deepseek_ai, @huggingface (SmolLM)
#LocalLLM #WebGPU #BuildInPublic
The desktop app is native (rust + tauri) and QR Code sharing for AI token is neat, love it really much, I may hop on product hunt soon...
Now it's not only for Open source projects of course, you can really give away AI tokens (Whatever the provider) to anyone that of course has moochy app desktop installed.
I am planning to team up with some compute provider to extends some features ideas I have in mind too.
DEJ retweeted
Want to learn System Design?
This website teaches it INTERACTIVELY, not through boring theory.
Explore it, learn it, and thank me later. 👀
🔗 designed-to-scale.vercel.app…
DEJ retweeted
Changes to the structure of data are easy to catch with tests.
But the meaning of data can change, even if other fields remain the same.
Here, Yemi goes over schema and semantic drift, and how changes to data types, timestamps, & more can silently corrupt data.
freecodecamp.org/news/schema…
DEJ retweeted
it's always funny to hear that JK Rowling is less racist than her fans
EXCLUSIVE: Following intense online backlash to Paapa Essiedu's casting as Severus Snape in HBO's upcoming 'HARRY POTTER' series, a bizarre narrative workaround was reportedly pitched to J.K. Rowling-which she immediately rejected!
Unconfirmed insider reports claim creatives floated an idea for Essiedu's Snape to use dark magic or Polyjuice-style transfiguration in the very first episode to permanently transform into a pale, book-accurate white man-an idea Rowling shut down outright as the series prepares for its December 25, 2026 premiere.
Readers added context they thought people might want to know
This is from @AnarchyCrave, which identifies as a parody and satirical news account in its profile bio. No credible reports support claims of any insider pitch to J.K. Rowling for a magical workaround on the Snape casting.
x.com/NebsGoodTakes/…
variety.com/2026/tv/news/h…
DEJ retweeted
if you want to re read this book again to make get rid of brain rot than just go through chapters 3, 6, 7, 8, 11 and 12
I always get something new to learn :D
AO VIVO COM IMAGENS: REESTREIA DE RONALDINHO GAÚCHO | RAVENNA FC 1913 X FORLI FC | ITATIAIA ESPORTE nitter.gq/i/broadcasts/2peleWRQk…
DEJ retweeted
When I started learning agentic AI, I kept running into three names: LangChain, LangGraph, LangSmith. They sound almost the same. What does each one actually do? Do you need all three?
Let me clear the confusion for you in 30 seconds.
>> 𝗜𝘁 𝘀𝘁𝗮𝗿𝘁𝘀 𝘄𝗶𝘁𝗵 𝗮𝗻 𝗟𝗟𝗠.
It can answer questions, write code, and summarize documents.
But it can't reach your database, call your APIs, or remember last week's conversation on its own.
So developers wrapped code around it: prompts, document retrieval, API calls, output parsers. Everyone ended up writing the same glue code, and switching models meant touching the whole app.
>> 𝗟𝗮𝗻𝗴𝗖𝗵𝗮𝗶𝗻 𝗳𝗶𝘅𝗲𝘀 𝘁𝗵𝗶𝘀.
It gives you ready-made building blocks to connect models, tools, retrievers, and prompts into one working app.
Then we see the problem
- A simple chain works when every step follows the last one.
- Real agents don't behave like that.
If a tool fails, the agent must retry.
It has to pick between two paths.
It needs a human to approve an action.
Or the app crashes halfway through a long task.
Handling all that with loops, flags, and custom retry code gets ugly fast.
>> 𝗟𝗮𝗻𝗴𝗚𝗿𝗮𝗽𝗵 𝗳𝗶𝘅𝗲𝘀 𝘁𝗵𝗶𝘀.
You build workflows where steps can loop, branch, share state, and resume after a failure.
Think of it as a stateful workflow engine for AI agents.
Now you ship it.
Real users are on it.
Then someone says, "The agent gave me a wrong answer."
Your logs look healthy. But what actually went wrong?
- Was the retrieved context bad?
- Did a tool return wrong data?
- Did the model ignore your instructions?
- Did your last prompt change break something?
- And what did that one request cost?
Normal app logs won't tell you.
>> 𝗟𝗮𝗻𝗴𝗦𝗺𝗶𝘁𝗵 𝗳𝗶𝘅𝗲𝘀 𝘁𝗵𝗶𝘀.
It traces each run, shows prompts and tool calls, tracks latency and token usage, and lets you test outputs against datasets before you roll out changes.
Think of it as observability plus evaluation for LLM apps.
𝗧𝗼 𝘀𝘂𝗺𝗺𝗮𝗿𝗶𝘀𝗲
LangChain → connects LLMs to tools, APIs, and data
LangGraph → manages state, loops, retries, human approvals
LangSmith → traces, debugs, and evaluates quality
You don't need all three on day one.
- Start simple. Add LangGraph when your workflow needs state and branching.
- Add LangSmith when you need to see what's happening in production.
Don't try to learn every framework. Learn the problem each tool solves.
DEJ retweeted
Snowflake and Databricks are becoming the SAME platform. The last 12 months have made that increasingly obvious.
Snowflake started as a cloud data warehouse.
Databricks started as a Spark-based data engineering and machine learning platform.
But look at what both platforms offer today:
1. Native dbt and SQL transformations
❄️ Snowflake: dbt Projects on Snowflake, Workspaces, and enhanced dbt DAG lineage.
🧱 Databricks: dbt on Databricks, dbt tasks in Lakeflow Jobs, and Databricks SQL.
2. Declarative data pipelines
❄️ Snowflake: Dynamic Tables, including custom incrementalization.
🧱 Databricks: Lakeflow Spark Declarative Pipelines, materialized views, and streaming tables.
3. Semantic layers
❄️ Snowflake: Semantic Views, Semantic Studio, and semantic view materializations.
🧱 Databricks: Unity Catalog Metric Views.
4. AI-powered analytics agents
❄️ Snowflake: Cortex Analyst, Cortex Search, and Cortex Agents.
🧱 Databricks: Genie Agents, Genie One, and Agent Bricks.
5. AI coding agents
❄️ Snowflake: Snowflake CoCo, including desktop, CLI, and scheduled automations.
🧱 Databricks: Genie Code, integrated into notebooks, SQL editors, pipelines, and dashboards.
6. AI functions inside SQL
❄️ Snowflake: AI_EXTRACT, AI_FILTER, AI_AGG, and AI_PARSE_DOCUMENT.
🧱 Databricks: ai_extract, ai_classify, ai_parse_document, and ai_query.
7. Infrastructure as code
❄️ Snowflake: Database Change Management (DCM) Projects.
🧱 Databricks: Declarative Automation Bundles.
8. Data governance and observability
❄️ Snowflake: Horizon Catalog, column-level lineage, and Data Quality Monitoring.
🧱 Databricks: Unity Catalog, column-level lineage, and Data Quality Monitoring.
9. Open table formats
❄️ Snowflake: Apache Iceberg support, Horizon Catalog interoperability, and Delta Direct.
🧱 Databricks: Delta Lake, Delta UniForm, Apache Iceberg interoperability, and Unity Catalog.
10. Low-latency analytics and applications
❄️ Snowflake: Interactive Tables, Interactive Warehouses, and Snowflake Postgres.
🧱 Databricks: Lakebase, Databricks SQL, and Lakehouse Apps.
Five years ago, the distinction was relatively straightforward:
Snowflake was for SQL analytics.
Databricks was for big data engineering and machine learning.
In 2026, that distinction is becoming increasingly difficult to make.
Both platforms are competing to own the ENTIRE data lifecycle.
The architectures are still different.
Databricks remains more deeply rooted in Spark, Python, streaming, and open-source data infrastructure.
Snowflake remains more deeply rooted in SQL, managed warehouse experiences, and enterprise analytics.
But now if you squint, you can see the features are basically the same! The vibe coding era will make the convergence even clearer in 2027!
This convergence has huge implications for analytics engineers.
The value of knowing which buttons to click in Snowflake or Databricks is declining!
The value of understanding data modeling, incremental processing, business semantics, data quality, and reliable system design is increasing.
Because those concepts transfer across both platforms.
In 2027, I think the best analytics engineers will be platform-aware but fundamentally platform-independent.
What do you think?
Will Snowflake and Databricks eventually become functionally interchangeable, or will their architectural differences always matter?
Sales close for my AI Analytics Engineering Boot Camp in just a few days! Book a discovery call with me today at DataExpert.io
dataexpert.io/
Mon ami ton contrôle pied droit passe pied gauche systématique comme si tu apprenait à utiliser tes jambes ne va pas le faire
🚨 Marc BERNAL 🇪🇸 sur son AMITIÉ avec Lamine YAMAL 🇪🇸, née au Barça à 7 ANS. 💙❤️
🗣️ « Je connais Lamine depuis l’âge de 7 ans.
On a joué ensemble jusqu’aux cadets A, puis il est monté chez les juniors A.
(…) On s’entend SUPER BIEN. Après autant d’années à jouer ensemble, forcément, une amitié se crée. »
🎙️ @partidazocope
112
DEJ retweeted
If you are a Data Engineer, here are 4 deep-dive articles worth saving.
Real experiments. Real numbers. Spark UI observations. And lessons you can use when things go wrong in production.
Save this post. You might want to come back to these.
- What happens when you read 40 Billion Records with Databricks Disk Cache?
medium.com/@sumitmittal-tren…
- How did AQE reduce a Spark job from 45 minutes to 21 minutes?
medium.com/@sumitmittal-tren…
- Parquet is already fast. So why does Delta Lake exist?
medium.com/@sumitmittal-tren…
- Why did 29,200 small files become a performance problem for Spark?
trendytech.in/the-small-file…
And the next experiment?
A Spark join that refused to get faster even after adding more compute.
That's what I'll break down next.
Until then, what production Spark problem should I experiment with next?
DEJ retweeted
You run the infra of 4 clients. On the 2nd, each asks: what did my platform cost last month?
One workspace per client: its own sources, members, budgets. One click to switch, nothing crosses over. Four answers, no spreadsheet.
👉 fivemetrics.io
#buildinpublic
Remercions le Seigneur d’avoir sauver celui qui a perdu guerre
DEJ retweeted
Je connais au moins 4 personnes qui ont perdu leur taff pcq la préfecture a tardé à leur renouveler leur titre de séjour c trop grave
Le Conseil d’État donnait 6 mois à la France pour corriger les dysfonctionnements de la plateforme des titres de séjour.
@NunezLaurent avait 6 mois pour agir. Mais il n'a rien fait de concret.
6 mois plus tard, des milliers de personnes vivent encore à la merci d’un papier.👇










































