koded

2025 wrap for a first-year CS student: Built products, won hackathons, leveled up in backend, AI, Web3, and DSA, all while keeping first-class grades & playing basketball. Thread 🧵👇

@MKingssAI @Lil_Tee18 @LASUOfficial check their VLUP 😭

@Lil_Tee18 @LASUOfficial I had to go check 🥲😅

@LASUOfficial Why does your website look vibecoded by MetaAI?

@Ifihan_ @Lil_Tee18 right when i was just learning to do it with Golang

I've been building BlazeRPC — an open-source Python library that lets you serve ML models over gRPC with minimal setup. Think FastAPI, but for gRPC. It handles proto file generation, adaptive batching, streaming, dependency injection, health checks & more. Works with PyTorch, TensorFlow, and ONNX out of the box. ⭐ Star it, contribute, or just check it out. Feedback and bug reports are very welcome! GitHub: github.com/Ifihan/blazerpc Docs: ifihan.github.io/blazerpc/

@Lil_Tee18 crazy, pele, there should be an extension to save code screenshots. I’ll check for you

My chromebook crashed today I was waiting till tonight to commit all my changes across diff projects only for it to crash and require a powerwash, no backup nothing

@Lil_Tee18 shall we invent the wheel? make we collab

I ran a controlled long-context experiment on an LLM to find what actually breaks retrieval. Results: • Baseline: 98.8% • 4k Random Noise: 97.5% • 1k Semantic Noise: 82.5% Conclusion: Context length isn't the primary failure, semantic similarity is. (1/4)

@Lil_Tee18 so it’s kind of like too many “RAG components(the documents), cause noise in the long run..?

The "Why": Failure occurs because semantically similar distractors cause Retrieval Interference. The model doesn't just "forget"; it gets confused by noise that looks like the target "needle." Semantic noise is ~4x more damaging than random noise, even at 1/4 the volume. (2/4)

@ennycodes samsung browser is wild 😂

frontend is easy

@platinum_on_top i thought we just import it with <link>

@platinum_on_top you learnt jQuery keh? Senior dev

Has to be jQuery Learned it, and never used it again.....

@platinum_on_top real

None, choose your mental health and open a farm instead.

@elonmusk he worked really really hard also,

19 years ago

Didnt get no sleep last night, Im late But time will tell 💚👻

@gabeciii @ennycodes read it as django less than 3; as about to ask some things buh never mind <3 same thing as a heart emoji

@ennycodes Django <3

I use express, you?

@ennycodes django 💚🤭

@SuperSisi woody dont fall 🥺

Men: We aren’t all just big kids Also men:

@MKingssAI @uniben_sug @SUG_UNIBEN @Uniblogmedia my boy 💚

I didn't know I was trending after the win 😂🙂‍↔️

Yakubu manage!!!! #NewProfilePic

@honour_can_code Agentic Startups

pitch your startup in just 5 words i’m listening

@dxbolaji DM, let’s talk 😊💚

@coder0214h I don’t even know what else to bomb again

i hate the fact that I have zero motion. please pmo