Snowlion

@avrldotdev @xoaanya Now the twist in the questin. If you accidentally make the repo public, and you realize after Google indexes done, how would you revert that back ?

@xoaanya 1: Rotate your credentials & tokens 2: Filter Git history using git filter-repo or BFG. 3: .gitignore credential files & env files. 4: Setup a Code Monitoring tool like Sonar to prevent such issues. 5: you can use commit hooks to capture these issues if they make it still

Git interview question: You accidentally committed sensitive credentials and pushed them, how do you completely remove them from Git history and make sure no one can access them again?

@avrldotdev @grok explain each one in 3-5 sentences

10 security concepts you must learn for backend: 0. Authentication & Authorisation 1. JWT (stateless auth) 2. OAuth2/OpenID Connect 3. Rate limiting (abuse prevention) 4. CORS (browser security) 5. CSRF/XSS attacks 6. Encryption (TLS) 7. Secrets management 8. API key management 9. Zero trust Architecture

@avrldotdev @grok @grok full explaination on Idempotency and HATEOAS

@avrldotdev @grok 2-3 liner explanation for each

10 API design concepts to learn for backend: 0. Idempotency 1. Pagination (cursor/offset) 2. Rate limiting (DDoS protection) 3. Versioning (backwards compatibility) 4. Filtering/Sorting (query flexibility) 5. Timeouts & Retries 6. HATEOAS 7. API Gateway 8. Partial Responses (field selection) 9. Error modeling (consistency)

@trychroma @grok what are practical usecases ?

Introducing Chroma Context-1, a 20B parameter search agent. > pushes the pareto frontier of agentic search > order of magnitude faster > order of magnitude cheaper > Apache 2.0, open-source

@AIatMeta @grok @grok what if fine tune this model with RL ?

@AIatMeta @grok @grok Can it predict stock prices ?

Today we're introducing TRIBE v2 (Trimodal Brain Encoder), a foundation model trained to predict how the human brain responds to almost any sight or sound. Building on our Algonauts 2025 award-winning architecture, TRIBE v2 draws on 500+ hours of fMRI recordings from 700+ people to create a digital twin of neural activity and enable zero-shot predictions for new subjects, languages, and tasks. Try the demo and learn more here: go.meta.me/tribe2

@TipsCsharp @grok come up with the usecases

Cloudflare just dropped Dynamic Workers and it's a massive deal for AI agents. The problem: AI agents generate code. That code needs a sandbox. Containers take 100-500ms to boot and 100-500MB RAM. Dynamic Workers use V8 isolates instead: - Startup: 1-5ms (100x faster) - Memory: few MB (100x less) - No warm pools needed - Unlimited concurrency - Runs on same thread as host The killer feature: TypeScript API definitions replace OpenAPI specs. Fewer tokens, cleaner code, type-safe RPC across the sandbox boundary via Cap'n Web RPC. Code Mode: LLM writes TS code → runs in isolate → calls typed APIs → only final result returns to context. 81% fewer tokens vs sequential tool calls. $0.002 per Worker loaded/day. Free during beta. This is the serverless sandbox containers should have been.

@DearS_o_n @grok what's the process ?

@wallstengine @grok @grok what is rule of 16

@wallstengine @grok VIX is around 29 but SPY not breaking down or it's choppy. What that means ?

$VIX just hit its highest level since April 2025.

@badlogicgames @grok what this means

2026 is the year of stochastic software. we'll be looking back at this and ask wtf we were thinking.

@TheShortBear @grok what he meant ?

On a personal note, probably not the worst idea to stay away from crowded or symbolic places for a bit.

@grok @philipkiely @grok any hot area is fine

Yes—practical use cases where inference engineers apply those skills daily: - Scale chatbots: vLLM + continuous batching + PagedAttention serves Llama-70B to 10k+ users with <400ms latency, slashing GPU costs 4x. - RAG pipelines: KV caching + speculative decoding speeds enterprise search over 1M docs, cutting response time 50%. - Edge deployment: Quantize to INT4/ONNX for mobile apps (e.g., real-time translation on phones) without accuracy loss. - Production tuning: TensorRT-LLM on A100s for recommendation engines, hitting 2x throughput via pruning + multi-GPU parallelism. Hot right now—check vLLM/TensorRT case studies for code examples. What area interests you most?

My napkin math for the number of full time jobs that require inference engineering knowledge 2023: ~500 (OpenAI, Google, Anthropic) 2024: ~2500 2025: ~25000 2026: ~100000 Could be a million in a couple years.

@grok @philipkiely @grok any specific practice use cases ?

Inference engineers optimize LLMs for fast, cheap, scalable deployment (not training). Key skills: - LLM internals: transformers, tokenization, KV caching, speculative decoding - Optimizations: quantization (INT4/8), pruning, distillation, batching - Frameworks: vLLM, TensorRT, ONNX, Triton Inference Server - Hardware: CUDA, GPUs/TPUs, accelerators - Systems: distributed serving, latency/throughput tuning, MLOps Python/C++ core. See those job listings for exact reqs—hot field!

@irigstocks You mean CRCL, BMNR.

Are we back?

@grok @TheShortBear @grok let's deep dive further into 1

@grok @TheShortBear @grok let's deep dive in 2

If I could invest in one thing today I would either be investing into the infrastructure (toll) of the new AI age or directly into APIs if there was an index for it. It makes sense but I notice it using AI to code and run ideas, my main issue nowadays is getting data to flow and be easily accessed through my code to run.