DEV Community • 2026-08-05 16:50

Google Expands Gemini Notebook With Visible Thinking Steps and Workspace Integration

Google is expanding its notebook-based research experience into Gemini Notebook, a more integrated product that connects the Gemini app, Google Search and Google Workspace. The significant new user-facing capability is the option for eligible subscribers to expand thinking steps during chats and notebook operations, offering more visibility into what the model is doing as it works. The change mov...

Hacker News: Front Page • 2026-08-05 16:50

Celld: Self-hosted, distributed Durable Objects

Article URL: https://github.com/denoland/celld Comments URL: https://news.ycombinator.com/item?id=49185430 Points: 15 # Comments: 0

DEV Community • 2026-08-05 16:49

Measuring real LAN speed from a phone with iperf3

Disclosure before we start: I build the native iPerf3 apps for iPhone, iPad, Mac, and Android mentioned near the end. Everything in this post also works with the free iperf3 CLI on two computers. The phone enters the picture when one endpoint has to move around the building. Speedtest answers one question: how fast is the path from your device to the ISP's nearest test server. Inside a home or of...

Hacker News: Front Page • 2026-08-05 16:47

Launch HN: HyperProbe (YC S26) – Agents that do read-only debugging in prod

Hi HN, this is Shailendra and Karan here. We are building a fast and safe way for coding agents to debug issues live in production.When prod breaks, it lets Cursor, Claude, and others drop virtual breakpoints or probes safely in your running code, and extract the exact variable values that logs don’t have.All this saves time and effort for engineers who’d otherwise dig through logs and traces or r...

DEV Community • 2026-08-05 16:46

What a Neural Network Actually Computes: From Token IDs to Matrix Multiplication

NLP models cannot process words directly, which is why tokenization[1] exists. Tokenization ends with a list of integers — token IDs like [30642, 1634, 318, ...]. That list is what actually gets fed into a model. This article answers why NLP models cannot process raw text directly. The answer is simple: NLP models consist of neural networks that perform matrix multiplication, which needs a list o...

Hacker News: Front Page • 2026-08-05 16:45

Why I'm leaving OpenAI to build telepathy

Article URL: https://naomibashkansky.com/blog/telepathy/ Comments URL: https://news.ycombinator.com/item?id=49185370 Points: 11 # Comments: 2

DEV Community • 2026-08-05 16:35

AbilityGuard v1.1: I Gave My Abilities API Monitor a Memory (and a Voice)

A few weeks back I published an article upon my latest AbilityGuard plugin so you could actually see what the WordPress Abilities API was doing on your site — which abilities were registered, which ones were risky, and what ran when nobody was watching. That part worked. But pretty quickly I ran into a problem I should've seen coming: a monitor that only shows you the last 100 log entries and mak...

DEV Community • 2026-08-05 16:33

From the Lab Bench to the Codebase: What Scientific Research Taught Me About Debugging

For over five years, my "bugs" weren't in code. They were in DNA sequencing runs that came back with unreadable peaks, PCR reactions that refused to amplify, and antioxidant assays that gave inconsistent results across replicates. I have a background in Biotechnology and a Master's in Chemical Biology. I spent years in research labs analyzing experimental data, documenting processes, and trying t...

DEV Community • 2026-08-05 16:30

LLD Data Structures in Design Context: How to Recognize Trie Problems Before Writing Code

"Experienced engineers don't recognize a Trie because someone mentions search. They recognize it because the business revolves around discovering values from partial input." So far in this Trie mini-series, we've learned: why some software problems require prefix matching instead of exact lookup, how a Trie organizes data around shared prefixes, and why it naturally powers auto-complete and s...

Lobsters • 2026-08-05 16:28

C++26: #embed

Comments

DEV Community • 2026-08-05 16:26

Micro-compaction: amortizing context compression in agent loops

If you've run a long agent session in any framework, you've hit the wall: context fills up, the framework decides it's time to compact, and everything stops while a summarizer chews through the whole transcript. In hermes-agent that batch compaction fires around 80% of the context threshold, and on a long session the pause is measured in minutes. The agent isn't stuck, it's doing necessary work, b...

Node.js Blog • 2026-08-05 16:25

Node.js 26.7.0 (Current)

DEV Community • 2026-08-05 16:21

One Long Prompt Shouldn't Freeze Everyone's Tokens: Prefill/Decode Disaggregation

An LLM request is two workloads in a trench coat — a heavy, bursty prefill and a stream of tiny latency-sensitive decodes. Running them on the same engines lets one big prompt stall everyone. Splitting them fixes it. TL;DR: Every LLM request is two very different jobs. Prefill reads the whole prompt — heavy, bursty, and slow for long contexts. Decode then emits tokens one at a time — tiny, but la...

DEV Community • 2026-08-05 16:21

Your Agent Has a Bug You Can't Reproduce. Here's How to Catch It.

Deterministic simulation testing drives every fault, clock, and random choice from one seed — so a flaky, once-in-production agent bug becomes a reproducible artifact you can shrink to one line. TL;DR: The worst agent bugs only appear under a specific interleaving of faults — a tool fails right after a side effect, a retry fires, and money moves twice. Happy-path tests miss it, and when it hits p...

DEV Community • 2026-08-05 16:21

Stop Making Your Agent Wait: Branch Prediction for Tool Calls

Speculative tool execution guesses the next tool while the model is still reasoning, runs it in parallel, and hides the latency — discarding the guess when it's wrong. TL;DR: An agent's think→act→observe loop is serial, so it burns a huge share of wall-clock sitting idle while a tool runs. Speculative tool execution borrows branch prediction from CPUs: a cheap predictor guesses the next tool call...

DEV Community • 2026-08-05 16:20

The Agent That Answers Before You Ask

Sleep-time compute: split the budget so a background worker does the predictable thinking while idle — and the user waits far less when they finally ask. TL;DR: Most agents only think when a request arrives — the user waits and you pay full latency every time. But between sessions there's idle capacity, and many queries are predictable variants of past ones over context that barely changes. Sleep...

DEV Community • 2026-08-05 16:19

Stop Streaming Tools Through Your LLM

The 2026 shift from tool-calling to Code Mode: let the agent write one script instead of narrating fifty tool calls — and watch context tokens drop ~99%. TL;DR: The classic agent loop loads every tool definition into the context window and pipes every intermediate result back through the model. Connect a few dozen tools and the context is full before the user even speaks. The 2026 move — Code Mod...

Hacker News: Front Page • 2026-08-05 16:19

Discovery Loop

Article URL: https://www.discoveryloop.com/ Comments URL: https://news.ycombinator.com/item?id=49184960 Points: 4 # Comments: 0

DEV Community • 2026-08-05 16:18

Más allá de @JavascriptInterface: Diseñando un ORM SQLite Reactivo Nativo para Android WebViews

: Más allá de @JavascriptInterface: Diseñando un ORM SQLite Reactivo Nativo para Android WebViews El Problema que Todos Ignoran Si alguna vez has construido una aplicación híbrida para Android combinando Java/Kotlin con un WebView, conoces el dolor de las complejidades en la comunicación entre ambos mundos. El flujo de trabajo habitual es frustrante: Definir un modelo en Java...

Hacker News • 2026-08-05 16:15

Show HN: I made alchemical-cosmological PCB badges

Comments