📣 Tune in for #LoFi/42 on Tuesday at 8 AM PDT (October 27).
1️⃣ Steven @liechti_dev: Rust, iroh and CRDTs in a native workspace
2️⃣ Kushagar @KushagarGoel: Bring local-first to your backend
3️⃣ Schalk Venter: Local-first without the commitment
🧵👇
⚡️ Introducing Streamgres : the sync engine for Multiplayer AI
Being built as an extension to postgres, all in Rust.
Learn more in our session at SF Tech Week on Oct 7
Register here : https://t.co/9sCPe8HGuR
🚀 Introducing XOR
An open source, multimodal Jev-like decision model. Being developed with data sovereignty & enterprise grade in mind.
❶ Multimodal: great for text + visual classification
❷ Much bigger 260k context window
❸ Qwen3.6 based
https://t.co/hId0r3iRXx
↓↓↓ See it in action
Future Devs will do everything but development
It was about doing QA as LLM has written code, now suddenly everyone is interested in finding how good agent is performing on not sure what metrics
Looking for online HR courses 😂😂
Have any one of you made any mistakes??
Then if My agent for once give wrong answer why is it a bad agent☹️
In defence it learned from human data itself😅
Our brain will think about something when you try to not think about it
When i say “not apple” you will still think about apple🍎
Does this happens with LLMs as well??
#aiPsycology
As Samay Raina you say
Ek clean comic kabhi harami nahi ban sakta but ek harami kabhi bhi clean comic ban sakta h
But we say
Ek vibecoder kabhi bhi Developer ni ban sakta but ek Dev kabhi bhi vibecoder ban sakta h 😎
#developer
𝗔𝗜 𝗺𝗮𝗱𝗲 𝗺𝗲 𝘁𝗵𝗲 𝗺𝗼𝘀𝘁 𝗽𝗿𝗼𝗱𝘂𝗰𝘁𝗶𝘃𝗲 𝘂𝗻𝗽𝗿𝗼𝗱𝘂𝗰𝘁𝗶𝘃𝗲 𝗗𝗲𝘃
Last few month I have made more than 20 projects
Still it doesn't feel productive
Solve hard problems because building same state of art repeatedly doesn't help anybody but LLM giants
#AI
You're destined to fail if you're building for today. By the time you ship, the world will have moved.
2 years from now, AI will be nearly free, compute will be on-device. Distribution, speed, and unique data will matter more than anything else. Start there.
#AI#Buildit
Most engineering decisions are guesses
You don’t know
• future traffic
• real bottlenecks
• user behavior
Yet people over-engineer to “future-proof”
Good engineers don’t
They pick the simplest thing
that can change later
#Engineering
Anyone can build a “complex” system
Add Redis, Kafka, microservices, Kubernetes
Your architecture diagram looks amazing
6 months later:
You’re debugging distributed failures and tracing logs
Good engineers don’t just ask: Can we build this?
They ask: Do we want this?
#Engineer
AI can design a system in seconds:
Microservices
Redis
Kafka
Vector DB
Event-driven architecture
Looks impressive.
But it doesn’t know your users, traffic, or team size.
The real skill now isn’t knowing tools.
It’s knowing when not to use them.
#Engineering
🔥 Fastest way to fail a system design interview:
We used Kafka.
Interviewer: Why?
If you can’t explain the bottleneck, you don’t understand the tool
Great answers sound like:
File store → slow writes → indexing → Postgres.
Don’t list tools.
Show reasoning.
#SoftwareEngineer
You’re not learning engineering but vocabulary
Knowing:
• Microservices
• Kafka
• RAG ≠ Understanding systems.
If you remove the tools, can you explain
•Why it’s structured that way
•What trade-offs exist
•What breaks under load
Buzzwords feel smart.
Depth feels painful
You don’t need Redis.
If your app has:
• 3 users
• 20 requests/day
• ₹0 revenue
You don’t need:
• Caching
• Microservices
• Event streaming
You need a file.
Start simple.
Scale when it hurts.
Complex ≠ Smart.
#Engineering
Consistency or Availability.
Pick two. Suffer quietly.
Then Google Spanner showed up and said:
“What if we engineer around the speed of light?”
Global ACID. Atomic clocks. Planet scale.
Did it beat distributed systems or just outspend physics? 🚀
#DistributedSystems#Database
Your database is a very expensive to-do list
If you use Cassandra / ScyllaDB, you’re not on a B-Tree.
You’re on an LSM Tree.
Not a data structure, a strategy:
• write to memory
• flush sorted files
• merge later (compaction)
Write fast. Clean later.
#Databases#SystemDesign
How a hash map in Haskell is different from a hash map in Python
In Python, updating a hash map mutates it in place.
In Haskell, maps are immutable.
Create a new version without copying everything
How? 𝗛𝗮𝘀𝗵 𝗔𝗿𝗿𝗮𝘆 𝗠𝗮𝗽𝗽𝗲𝗱 𝗧𝗿𝗶𝗲
Only the updated branch is copied