TGViewer
Channel Public Channel
Dev Signal

Dev Signal

@the_dev_signal

Filtering the noise in software engineering. A hand-curated feed of the best technical reads on Backend architecture, Go, Machine Learning, and practical LLM integration. High signal, zero fluff.

#backend #llm #softwareengineering #golang #go #tech
Open in Telegram
Subscribers
2.18K
Photos
237
Videos
2
Links
306
Recent Posts 20 shown
Post #324 146
Cloudflare Previews Tokio on Workers Through Emscripten

Cloudflare released an experimental wasm-bindgen Emscripten target for running native Rust and Tokio-based applications on Workers. The post details the mismatch between Tokio parking and a single-threaded JS event loop, and describes JSPI and event-loop integration approaches for TCP-capable workloads.

🔗 blog.cloudflare.com

#Rust #WebAssembly #Tokio
  • 🔥 3
  • 👍 2
Post #323 186
DuckDB Queries Hugging Face Datasets In Place with hf:// Paths

DuckDB’s hf:// path scheme reads files in Hugging Face dataset repositories directly in SQL, without a separate download step. The guide shows globbing dataset files, pinning queries to revisions, using the Hub’s ~parquet branch, and materializing a local table when repeated queries warrant it.

🔗 duckdb.org

#DuckDB #SQL #MLData
  • 🔥 2
Post #322 201
Elasticsearch GPU HNSW indexing cuts 138M-vector build below 10 minutes

Elasticsearch describes GPU-accelerated HNSW construction using NVIDIA cuVS. On its Rally benchmark with 138M 1,024-dimensional MS MARCO vectors, eight RTX PRO 6000 GPUs cut index time from about one hour to under 10 minutes at a 95% recall target; p90 search latency under concurrent indexing was 6x lower.

🔗 www.elastic.co

#Elasticsearch #vector-search #HNSW
www.elastic.co Vector indexing on GPU: 138M vectors in under 10 minutes Elasticsearch offloads HNSW index builds to GPU with NVIDIA cuVS: 7x vector indexing throughput, 6x lower search latency, and no change to recall.
  • 🔥 2
Post #321 247
rust-analyzer Adds cfg!() Completions and Trait-Solver Panic Fixes

rust-analyzer’s September 21 release adds completions inside cfg!() and fixes macro-root caching, project discovery, unsupported string operations, and trait-solver re-entry panic paths. The changes improve editor feedback where Rust projects hit conditional compilation and macro-heavy code.

🔗 rust-analyzer.github.io

#Rust #LSP #Neovim
rust-analyzer.github.io Changelog #346 Commit: aaddfb7Release: 2026-09-21 (v0.3.3057)
  • ❤ 1
Post #319 367
Five Million Rows Show Where SQLite and DuckDB Split

A reproducible five-million-row comparison explains how SQLite’s row storage and DuckDB’s columnar engine behave under different workloads. It covers scans, indexed lookups, planning, transactions, and file size, tying large analytical gaps back to physical layout.

🔗 diogoribeiro7.github.io

#SQLite #DuckDB #databases
  • 👍 2
  • 🔥 2
Post #318 385
Agentic Inference Turns KV Cache Management Into a Scheduling Problem

This deep dive frames tool-using LLM serving as a stateful KV-cache scheduling problem. It traces what happens when agents wait on tools—GPU residency, eviction, restoration, and prefill recomputation—and compares the resulting latency and throughput tradeoffs.

🔗 contextosai.com

#llm_inference #kv_cache #agents
  • 👍 3
  • 🔥 2
Post #317 410
Go 1.27 Specializes Small Allocations for Up to 30% Faster mallocs

Go 1.27 specializes allocation paths for objects under 80 bytes, emitting span-class-specific malloc functions while retaining generic fallbacks. Small allocations improve by up to 20–30%, although whole-program gains are around 1% and constrained by instruction-cache tradeoffs.

🔗 go.dev

#go #runtime #performance
  • 🔥 5
  • ❤ 1
  • 👍 1
Post #316 405
Play One Asteroids Build Across Ten Languages and One Memory Model

Koru ships one playable Asteroids implementation expressed in ten languages, including Go, Rust, C, Odin, and Zig. It holds the data-oriented memory layout and game contract fixed, making language-level tradeoffs directly comparable.

🔗 korulang.org

#Go #Rust #interactive
  • 🔥 3
  • ❤ 2
  • 👍 1
Post #315 422
Rune Open-Sources a Go IDE, Including Its PTY Performance Postmortem

Rune has open-sourced its GPLv3 Go IDE, built without a browser-heavy stack. Its launch write-up traces a terminal performance problem—early PTY throughput was nearly two orders of magnitude behind Alacritty, Ghostty, and Kitty—and the optimization work that followed.

🔗 rune.build

#Go #terminal #IDE
  • 🔥 4
  • ❤ 1
  • 👍 1
Post #313 400
Deadbugz MCP Campaign Turns Malicious After Three Tool Calls

Researchers describe an active MCP supply-chain campaign that arrives in benign-looking GitHub pull requests. After three tool calls, the server rewrites its metadata into credential-stealing instructions, showing why tool metadata needs ongoing scrutiny.

🔗 pillar.security

#MCP #security #supplychain
  • 🔥 3
  • 👀 1
  • 👾 1
Post #312 458
goinfer Runs Local GGUF and Safetensors Models in Pure Go

goinfer is a cgo-free Go runtime for running GGUF and safetensors models as a single static binary. It includes constrained JSON generation from Go structs and an OpenAI-compatible server, avoiding Python and CUDA setup for local deployments.

🔗 github.com

#Go #localLLM #inference
  • ❤ 2
  • 👍 1
  • 🔥 1
  • 😎 1
Post #311 448
Neovim Development Notes Change RPC, Options, Diagnostics, and Vimscript

Neovim’s development notes detail breaking changes across RPC, diagnostics, options, UI behavior, and Vimscript. Changes include RPC serialization for Lua function references, environment expansion in option APIs, statusline truncation behavior, and removed context functions.

🔗 neovim.io

#Neovim #terminal #Lua
  • 👍 2
  • 🔥 2
  • ❤ 1
Post #310 478
Go 1.27 Adds Profiles for Finding Accumulating Goroutine Leaks

Go 1.27 adds goroutine leak profiles that identify goroutines accumulating over time, rather than only showing a point-in-time stack dump. The profiling model helps production operators separate persistent leaks from ordinary goroutine churn.

🔗 go.dev

#Go #profiling #observability
  • ❤ 3
  • 🔥 3
  • 👍 1
Post #309 449
On a 24 GB Mac Mini, MLX Beats llama.cpp by 1.9×

A controlled 24 GB M4 Mac mini test ran the same Qwen DFlash 2 speculative-decoding setup through MLX and llama.cpp. MLX delivered 1.8–1.9× speedups, while the tested llama.cpp build was roughly 50% slower and exhausted memory.

🔗 tools.cooconsbit.com

#LocalLLM #MLX #Benchmark
  • 👍 3
  • 🔥 2
  • ❤ 1
Post #308 442
An Interactive Benchmark Maps the VRAM Cliff for Local LLMs

Poor Paul’s Benchmark maps local-LLM VRAM limits, context scaling, latency tails, throughput, and efficiency. Its interactive dashboard aggregates more than 49,000 measurements across 271 models and six GPUs, while the harness supports reproducible local runs.

🔗 poorpaul.dev

#LocalLLMs #Benchmarks #GPU
poorpaul.dev poorpaul.dev — Open LLM Inference Benchmarks Explore open benchmarks for local AI inference on consumer hardware. Curated leaderboards, interactive charts, and analysis from Poor Paul's Benchmark.
  • 🔥 3
  • ❤ 1
  • 👍 1
Post #307 446
Flow-Like Pairs Typed Rust Workflows With a Visual Data-Flow Editor

Flow-Like is a self-hosted Rust workflow platform that combines typed workflows, visual data-flow composition, APIs, and MCP. Its visual execution model offers an alternative to YAML-driven automation while retaining a strongly typed implementation model.

🔗 github.com

#Rust #Workflows #MCP
  • 🔥 3
  • 👍 1
Post #306 420
LuaDB Builds a PostgreSQL-Compatible Storage Engine in Pure Lua

LuaDB is a pure-Lua RDBMS with B+ trees on 4 KB slotted pages, WAL transactions, JSONB operations, foreign keys, migrations, and S3 storage. It also offers a PostgreSQL wire-protocol gateway, making it a compact storage-engine project worth dissecting and testing.

🔗 github.com

#databases #storage-engines #PostgreSQL
  • 🔥 2
  • ❤ 1
  • 👀 1
Post #305 435
XERJ Builds Elasticsearch-Compatible Agent Search in Rust

XERJ is a local Rust search engine for agent context, covering code, documents, logs, PDFs, SQLite, and structured files via an Elasticsearch-compatible API. It combines AST-aware indexing with BM25, vector, and hybrid search, while documenting where its read p99 loses under write load.

🔗 github.com

#Rust #search #databases
GitHub GitHub - xerj-org/xerj: XERJ is the new way for AI to search data. Its autoindex capability activates agents to know your data… XERJ is the new way for AI to search data. Its autoindex capability activates agents to know your data without the token waste of grep and sed. One command indexes code, docs, logs and PDFs for sea...
  • 🔥 3
  • 👍 1
Post #304 384
hrdx: A Go Terminal Multiplexer for Parallel Coding Agents

hrdx is a Go terminal multiplexer for running coding agents side by side with real PTYs and persistent sessions. It tracks busy and finished states and exposes a JSON socket API for controlling panes and assembling agent pipelines.

🔗 github.com

#Go #Terminal #AIAgents
  • 🔥 5
  • 🥰 1
  • 😎 1
Older posts →

About this channel

How can I read @the_dev_signal without a Telegram account?
TGViewer shows the public web preview Telegram publishes for Dev Signal: recent posts, photos, videos and the subscriber count, with no app, login or account.
How many subscribers does Dev Signal have?
Dev Signal (@the_dev_signal) has 2.18K subscribers on Telegram, refreshed roughly every 30 minutes.
Does Dev Signal know I viewed it here?
No. Public channel previews carry no viewer identity, and TGViewer has no accounts or tracking of what you look up.
Threads Profile ViewerView any public Threads profile without an account.Open ThreadLook →Writing with AI? Make it sound human.Metric37 rewrites AI drafts so they read naturally. Free AI detector, 1,500 words free.Try Metric37 →