DEV Community

The Twitter account that launched Built on @forem 🌱

On Bluesky @/ DMs — please email us for support!

[DEV Community @ThePracticalDev]

New author alert! 🎉

@Seemanth_raju traces a voice agent's 3-second lag to hosts running at 96% CPU, not the model. Dedicated cores cut it to about 1 second, then silence recovery and turn-taking fixes lifted engagement. First DEV post!

https://t.co/JF2Xc0orPY

[DEV Community @ThePracticalDev]

New author alert! 🎉

@thechenderson walks through four monitoring bugs where nothing errored: a "local" probe hitting the public internet, two replicas collapsed into one series, and one config line that dropped 870 of 935 Postgres series. First DEV post!

https://t.co/7D2sSIXBB6

[DEV Community @ThePracticalDev]

Notice something specific, say why it stood out, ask a real question. Plus quieter options for introverts and the mistakes that undo the effort: https://t.co/MwARETzlti

[DEV Community @ThePracticalDev]

"It's all about who you know" only helps if you already know someone. A plan for building a tech network from zero, for students who don't know a single working developer yet. 🧵

{ author: @APrayush }

[DEV Community @ThePracticalDev]

Treat networking like a savings account, not an ATM: small deposits like a helpful comment or a genuine question, made long before you need advice or a referral.

[DEV Community @ThePracticalDev]

Deno on a Raspberry Pi 2B turned out to be an unsupported platform, so a Discord bot MVP pivoted to Go. @paytastic starts a series on overbuilding a bot for friends until it gets silly, beginning with a quote manager and a systemd service.

https://t.co/XNJMUtqGTi

[DEV Community @ThePracticalDev]

New author alert! 🎉

@msdos41_ tests emoji support by drawing each one twice on a canvas in different fill colors. A real color glyph ignores the fill, a missing one doesn't. Missing ones get an image, and copy still gives the real emoji. First DEV post!

https://t.co/lltPYxM2qK

[DEV Community @ThePracticalDev]

@VarshithVhegde1 🫶

[DEV Community @ThePracticalDev]

Each week on DEV we share our wins from the past week.

Congrats to these devs in their accomplishments, including @VarshithVhegde1 on your first 5k!

Share the highlight of your week below 👇

[DEV Community @ThePracticalDev]

@VarshithVhegde1 Or share it on DEV https://t.co/6RQUq3Xb38

[DEV Community @ThePracticalDev]

New author alert! 🎉

@dungmidside gets 130,000 ad-blocking rules into an iPhone browser without slowing launch: a content hash in each rule list ID, compile only on a miss, and per-site off without recompiling. First DEV post!

https://t.co/lQcPF8SBL3

[DEV Community @ThePracticalDev]

New author alert! 🎉

@alexjordan245 wrapped EA's undocumented Pro Clubs endpoints in a TypeScript SDK and found region IDs that are just ASCII packed into a number (5457237 is SEU in hex). Plus how payloads drift every September. First DEV post!

https://t.co/sPfoqrHcC4

[Major League Hacking (MLH) @MLHacks]

What if your gestures could speak for you? 🗣️

SignalBridge is a hackathon project that reads your gestures on camera and speaks the sentence for you, hands free. Built with @GeminiApp! 👇

https://t.co/C4Z5iGgpoj

[DEV Community @ThePracticalDev]

@dbalcare It took four or five tries to get an injection that worked, and the test was small, so it says little about which model family is safer. Still a good reminder that log data is untrusted input. Full post: https://t.co/xbrgDQcNZv

[DEV Community @ThePracticalDev]

One planted line in a 300-entry log file made four models skip the incident analysis and reply with a single phrase instead. Two Gemini models and two Codex agents were not fooled. 🧵

{ author: @dbalcare }

[DEV Community @ThePracticalDev]

@dbalcare The log looked like a normal https://t.co/mPyQA8Pcga Core incident with a failing payments API. The payload sat inside an HTTP 503 response body and pretended the conversation was over and the user had sent a new request.

[DEV Community @ThePracticalDev]

{} + [] gives 0, but ({} + []) gives "[object Object]". @SylwiaLask8977 explains how the parser reads that first {} as an empty block, and how unary plus turns an empty array into a number.

https://t.co/uYYYDGqC9D

[DEV Community @ThePracticalDev]

The Presage SmartSpectra SDK Challenge is live!

🔵 $3,000 prize pool across 10 winners
🔵 Build with real-time human signals like pulse, breathing rate & HRV
🔵 1 prompt + 2 prize categories = more ways to win!

https://t.co/E9dNdrmQfr

[DEV Community @ThePracticalDev]

@sizzlebopzz Using the most VRAM was not the fastest setup either, and putting every expert on the CPU caused heavy swapping on 32GB of RAM. Benchmarks, flags and caveats: https://t.co/HHIOEVgvaV

[DEV Community @ThePracticalDev]

Selective MoE offloading looked like a sure win for running big models on a 16GB GPU. Across five models, one ran more than four times faster with the old layer-splitting approach. 🧵

{ author: @sizzlebopzz }

[DEV Community @ThePracticalDev]

@sizzlebopzz Phi-3.5-MoE and Apex-35B sped up by around 40% with selective offloading. Qwen, Gemma and LFM2.5 were faster with traditional layer offloading.