RitiK Mehra

To infinity and beyond!

The timeline is all glow-ups and discipline, but no one posts the part where they almost gave up twice before lunch.

I was testing content generation on my local Ollama setup and kept getting that same robotic, context-blind tone. Felt useless for anything beyond demo level. So I built a tiny layer on top: - session memory - project context - trend feed from X - and a lightweight retrieval buffer Once Ollama had history + relevance + what’s happening today, the output changed instantly. It started sounding like an actual dev who understands my workflow, not a stateless text generator. Still not perfect, it forgets long chains, and trend-weighting sometimes overpowers the actual intent, but this is the closest I’ve seen a small local model get to “aware.” Next I'll try to add a reflection step so it can evaluate its own output before sending it. Basically giving it a second thought process. Running things on your local is actually fun.

@Priyansh_31Dec @grok is this real?

Finally got a backpack after 19 months🥹

@griftersingh auto wala bar se wapis job p jara h

The swachchhta psyop is real

@KevinNaughtonJr then you'll never be him

you don't need to understand how memory management works to write good code

The worst feeling in tech? When you finally understand the bug... and realize you wrote it.

Looking for a backend dev role - been building scalable systems, APIs, and self-hosted projects for years. If your team is hiring, I'd love to connect. Portfolio: ritik-mehra.site

Fr!

@KevinNaughtonJr and here you are creating more data

why do software engineers get paid so much we literally just move data around

act first, ask questions never!

At this point Cloudflare and AWS feel like the internet's biggest single points of failure. One hiccup and everything collapses.

Survived the AWS outage and the Cloudflare outage. At this point, I’m more fault-tolerant than half the internet.

git repo: github.com/shubh73/devfol… thanks to @shubhporwal24! :)

To infinity and beyond!

@shubhporwal24 @mannupaaji Ahh forgot to mention Shubh paaji <3, thanks for amazing work!

@rtk_mehra @mannupaaji love seeing people use my portfolio. though asking paaji for a review on a fork is wild 😭 repo: github.com/shubh73/devfol… @mannupaaji this one might deserve a special review 👀

Dropped a video on fixes that you can make in your portfolio website to > Get more job opportunities > Get more freelance clients > Rank better on google Give it a watch

Cloudflare outage hit and my pirated stream instantly died. Perfect timing bro, universe be watching. #Cloudflare

@mannupaaji my pirated streaming site got stucked -_-

Every website that uses cloudflare was down But Cloudflare itself wasn't Maybe they're hosted on Vercel

I just read about cache seeding, and honestly didn’t realize how many big systems rely on it to stay fast. Most cache issues don’t happen during normal traffic - they show up during deploys, cold starts, or when a node restarts and everything misses at once. Cache seeding basically means warming the cache before users hit it, so the system doesn’t panic. A few interesting examples I found: Netflix actually pre-warms petabytes of cache before peak traffic. (blog.bytebytego.com/p/how-netflix-…) Uber built something called “CacheFront” that helps them serve 40M+ reads/sec by keeping important data warm. (uber.com/en-IN/blog/how…) The Amazon Dynamo paper also talks a lot about using warm caches for high availability. (allthingsdistributed.com/files/amazon-d…) What I like about this concept is how simple the idea is: load the important stuff ahead of time so your system doesn’t suffer later. Common ways teams do it: - warm the cache during deploy - have background tasks refresh hot keys - write-through caching - use shadow traffic to pre-fill - batch load expensive queries Kind of funny how a “simple warm-up” ends up being the thing that keeps huge systems from falling apart during spikes.

@aslamdev912 @arcanedotbuild Glad you liked it :)

@rtk_mehra @arcanedotbuild Man, awesome website. Loved it.

We are hiring for: 1.Full-Stack Web3 Engineer (React + Next.js + Node) 2.Backend Engineer (Node/Python + Web3 Integrations) 3.Mobile Developer (React Native + Wallet Integrations)

I’ve been diving deeper into distributed systems lately, and the more I read, the more I realize how misunderstood they are. It’s not just “multiple servers working together.” It’s a constant negotiation between latency, failures, trade-offs, and reality. In distributed systems: - Every node has a different view of the world - Clocks drift - Messages arrive late or out of order - Services fail silently - “Truth” becomes a matter of perspective You don’t design for perfection, you design for failure, because something is always failing somewhere. And yet… that’s the beauty of it. You start appreciating how much engineering goes into making things look “instant” and “reliable” to users. A few things that hit me lately: - Consistency isn’t free, you pay with latency - Availability isn’t guaranteed, you build it through redundancy - Scaling isn’t linear, it’s a mix of caching, sharding, and good guesses - Fault tolerance isn’t magic, it’s retries, backoff, replication, and monitoring - Global truth doesn’t exist, only eventual agreement Distributed systems look clean in diagrams… and then the real world enters with network delays, partial failures, thundering herds, and CAP theorem reminders. But despite all the chaos, they’re the backbone of everything we use today, from payments to streaming to social apps. The more I learn, the more respect I have for anyone who keeps these systems running. Distributed systems aren’t hard because of scale. They’re hard because of reality.