agentplace.io
Show HN: Agentplace, the tool we built to become a 20x companyhttps://agentplace.io/Garry Tan posted a video this week about what he calls 20x companies. The idea is that small teams beat incumbents many times their size by automating everything internally, not just one or two functions. This alpha is real, and I think it is underhyped tbh.But here's the honest part, building agents is hard, expensive and most of the scaffolding we make would be obsolete with a new labs release. at the same time models are not quite there and they are not keeping up with "coding is dead" and "knowledge workers are doomed". Pushing them beyond what they can actually do reliably just boils the ocean. And we don't need to.Staying in sync with Labs' release pace is already super mega photonic speed. So to sum up: we need to use what they are capable of, keep our options open and quickly improve our agents. How to do it w/o a dedicated person for AI innovation?That problem is what we kept running into at Agentplace.We wanted a builder that handles the full stack for an internal agent: backFeb 27, 2026 12:42 AM
news.ycombinator.com
Brave New Smart Phone Dependence World and Google Supporthttps://news.ycombinator.com/item?id=47173467tl;dr: I've spent countless hours with Google Support, but the support process has completely failed.I was originally offered a mail-in repair (RMA) to my old address in Germany, but I've left that country and cannot receive the package there. Maybe I am the first digital nomad with Google products? I don't know... :(Despite confirming my new location in Turkey, I was told by a supervisor that an Advance Exchange replacement cannot be shipped to me because the Pixel 10 was not launched in Turkey. But the supporter before told me to delete my payment profile from Germany. Now, what?!Google does not offer a loaner device, meaning I am left without a necessary smartphone for critical functions (payments, travel, communication) while the device is sent for a 7+ day repair (if I could even send it in). The issue is just a small crack which made the fingerprint sensor unusable. Apple can repair such things on the spot in the whole world...My health and livelihood are being severely impacted Feb 26, 2026 11:08 PM
duckui.com
Show HN: Duck-UI Embed – SQL-Powered React Dashboard Components with DuckDB-WASMhttps://duckui.com/play/Hi HN,I built @duck_ui/embed — a set of React components that turn SQL queries into interactive dashboards, powered by DuckDB-WASM running entirely in the browser. `npm install @duck_ui/embed @duckdb/duckdb-wasm` The idea: pass your data (arrays, CSV, Parquet, JSON, or a fetch callback), write SQL, and get charts, tables, KPIs, and filters — no backend required. What's under the hood:- DuckDB-WASM boots in the browser, data loads into in-memory tables - SQL queries run against real DuckDB (full SQL support, window functions, CTEs) - Parquet files use HTTP range requests — only fetches needed row groups - FilterBar auto-detects column types and injects WHERE clauses via subquery wrapping - Connection pool (max 4), LRU query cache (100 entries, 5min TTL) - ~95KB bundled (ESM), tree-shakeableComponents: Chart (line/bar/area/scatter/pie), DataTable (paginated, sortable, resizable columns), KPICard (with sparkline + comparison), FilterBar (auto or manual config), Dashboard (responsive grid Feb 26, 2026 9:16 PM
github.com
Show HN: pg_stream – incremental view maintenance for PostgreSQL in Rusthttps://github.com/grove/pg-streampg_stream is a PostgreSQL 18 extension that keeps materialized views current incrementally — no external infrastructure, no separate streaming pipeline.You define a stream table (a SQL query + a freshness bound) and the extension derives a delta query that processes only changed rows on each refresh cycle: SELECT pgstream.create_stream_table( 'regional_totals', 'SELECT region, SUM(amount) AS total, COUNT(*) AS cnt FROM orders GROUP BY region', '1m', 'DIFFERENTIAL' ); One INSERT into a million-row table → pg_stream touches one row's worth of computation. Query it like any regular table.How it works: - Trigger-based CDC captures row changes into buffer tables inside the same transaction (no wal_level = logical required) - A background worker walks a topological DAG of stream tables and fires refreshes in dependency order - The DVM engine (differential view maintenance, DBSP framework) rewrites your SQL into a delta query at definition time — JOINs, GROUP BY, CTEs, window functions, LATERFeb 26, 2026 7:17 PM
news.ycombinator.com
Show HN: A Vaadin Algebra and Calculus Solver Built with AI Assistancehttps://news.ycombinator.com/item?id=47116251Hi HN,I’d like to share a side project that has gradually become my main creative outlet: The Algebrator, a web-based algebra and calculus solver I built using Java, Spring Boot, Vaadin 24, and a symbolic math engine under the hood.It started as a personal attempt to revisit the math I loved in middle/high school, but it evolved into a multi-year design experiment in AI-augmented software engineering. I built and iterated on this app using a paired-programming workflow with LLMs (ChatGPT + GitHub Copilot), and the project ended up teaching me more about agentic AI development than anything else I’ve worked on.What it doesLets users type equations, inequalities, and expressions in a calculator-like UISolves algebraic equations, systems, trigonometry, calculus operations, and iterative “problem templates”Supports fraction/decimal modes, radians/degrees, comparison operators, user-defined functions, and multi-character variablesIncludes utility “extras” like prime generation, Fibonacci, rFeb 23, 2026 12:04 AM
news.ycombinator.com
Show HN: Vexp – graph-RAG context engine, 65-70% fewer tokens for AI agentshttps://news.ycombinator.com/item?id=47113273I've been building vexp for the past months to solve a problem that kept bugging me: AI coding agents waste most of their context window reading code they don't need.The problemWhen you ask Claude Code or Cursor to fix a bug, they typically grep around, cat a bunch of files, and dump thousands of lines into the context. Most of it is irrelevant. You burn tokens, hit context limits, and the agent loses focus on what matters.What vexp doesvexp is a local-first context engine that builds a semantic graph of your codebase (AST + call graph + import graph + change coupling from git history), then uses a hybrid search — keyword matching (FTS5 BM25), TF-IDF cosine similarity, and graph centrality — to return only the code that's actually relevant to the current task.The core idea is Graph-RAG applied to code:Index — tree-sitter parses every file into an AST, extracts symbols (functions, classes, types), builds edges (calls, imports, type references). Everything stored in a single SQLite file Feb 22, 2026 6:14 PM
deathwink.com
Show HN: Deathwink – Send messages to people after you diehttps://deathwink.comHey HN. I built Deathwink (https://deathwink.com) -- a web app that delivers your messages after you die.You write a message, record a video or audio message, or both -- attach whatever you want, add your recipients, and go live your life. Every 30 days, we check in. If you stop responding, after 90 days we assume you've died and deliver your messages.The idea came from a real place. In my early 20s, my best friend Sean and I came up during the first dot-com boom. We were broke, dumb, and inseparable. He died of an overdose a few years later. I never told him what his friendship meant to me—not because I didn't feel it, but because that's not what guys like us did. I wrote a memoir about it (How You Wish You Could Leave), and somewhere in the process of reliving all that regret, I realized I wanted to build the thing I wished had existed.On the technical side: it's serverless on AWS -- Lambda functions behind API Gateway, RDS, Angular frontend on S3/CloudFront, Cognito for auth, SES foFeb 17, 2026 4:18 PM
spacedaily.com
Flexible electronics reshape intelligent robot designhttps://www.spacedaily.com/reports/Flexible_electronics_reshape_intelligent_robot_design_999.htmlTokyo, Japan (SPX) Feb 08, 2026 With robots taking on more roles in daily life and industry, their evolution toward autonomous intelligence is being held back by rigid electronic hardware that cannot conform to complex, curved bodies. Traditional hard components are difficult to mount on non planar robot surfaces, which limits the integration of critical sensing, computation, and actuation functions that underpin perception, dFeb 11, 2026 11:57 AM
news.ycombinator.com
Tell HN: AI is all about the tools (for now)https://news.ycombinator.com/item?id=46742677It's like having a body. If you don't have a body, you can't do much.Tools that agents have access to are the agent's body. Without them, they can't do anything.For example, an agent without an ability to extract individual functions and check them in isolation, respecting the AST, has a hard time simply re-ordering and re-grouping the functions in a large file.Even a frontier model struggles with this. In the same way that you would struggle to navigate the ocean without a sextant, or struggle to join two pieces of metal without a welder. Without tools, intelligence doesn't get you much in practice.Of course - eventually agents will probably "craft their own tools", all the time, instantly. Many times they already do this, ad hoc bash/pythong/perl (or even sed or awk) scripts to do large replacements, or mechanical changes.But these kind of thigns are still blunt instruments. Not exactly Switch watches or tooling with some elaborate regex and string replacing. If you give AI better toJan 24, 2026 11:22 AM
news.ycombinator.com
Show HN: PolyMCP – a toolkit for MCP servers and agent integrationhttps://news.ycombinator.com/item?id=46632556PolyMCP is a framework for building and interacting with MCP (Model Context Protocol) servers and for creating agents that can use those servers as dynamic tools.Developing and managing MCP servers often involves several challenges:Exposing Python functions as discoverable tools requires repetitive boilerplate.Coordinating multiple MCP servers simultaneously is complex.Debugging tools during development is hard due to limited visibility into calls and outputs.Integrating agents with LLMs to automatically discover and invoke tools is usually ad-hoc.PolyMCP addresses these challenges by providing:Flexible tool exposure: Python functions can be served over HTTP, in-process, or stdio with minimal setup.Inspector dashboard: Monitor tool invocations, metrics, and test calls interactively in real-time.Agent integration: Support for multiple LLM providers and automatic tool discovery and invocation.CLI and workflow utilities: Streamline project setup, testing, and execution.PolyMCP reduces friJan 15, 2026 1:53 PM
spacewar.com
Terran Orbital to build satellite buses for SDA Tranche 3 Tracking Layerhttps://www.spacewar.com/reports/Terran_Orbital_to_build_satellite_buses_for_SDA_Tranche_3_Tracking_Layer_999.htmlLos Angeles CA (SPX) Jan 01, 2026 Terran Orbital has been selected by Lockheed Martin to supply satellite bus platforms for the Space Development Agency's Tranche 3 Tracking Layer within the Proliferated Warfighter Space Architecture. Under this program, Lockheed Martin will deliver 18 space vehicles that provide missile warning, tracking, and defense functions for SDA's Tranche 3 Tracking Layer. Terran Orbital will supplyJan 5, 2026 5:37 AM
spacewar.com
RTX radar selected to support autonomous X 62A fighter testinghttps://www.spacewar.com/reports/RTX_radar_selected_to_support_autonomous_X_62A_fighter_testing_999.htmlLos Angeles CA (SPX) Jan 01, 2026 Raytheon has received a U.S. Air Force contract to integrate its PhantomStrike radar on the autonomous X-62A Variable In-flight Simulation Test Aircraft, or VISTA, a modified F-16D Block 30 test aircraft upgraded with Block 40 avionics and configured as a machine-learning and software testbed. The X-62A VISTA functions as a dedicated platform for flight testing autonomy and artificial inteJan 1, 2026 11:20 AM
github.com
Show HN: Auto-Generated Captions with Twickhttps://github.com/ncounterspecialist/twickI am working on Twick Studio, an intuitive visual editor for video built on top of the Twick SDK, react toolkit for creating custom video editing experiences.Key capabilities include: - Multi-track, drag-and-drop timelines - Live preview of edits - Canvas editing with drag, resize, and rotate - Undo/redo and controls for text, video, audio, and images - Media utilities for metadata and thumbnail generationRecently added new cloud functions in the Twick SDK to auto-generate captions and export videos as MP4. You can watch a short demo here: https://youtu.be/zgRcJt8wxLkTry Twick studio: https://development.d1vtsw7m0lx01h.amplifyapp.comAuto-generated captions: https://development.d1vtsw7m0lx01h.amplifyapp.com/subtitlesDocs: https://ncounterspecialist.github.io/twick/docs/introSource code: https://github.com/ncounterspecialist/twickIf you find it useful, a on GitHub is appreciated.Dec 14, 2025 8:11 PM
spacedaily.com
SatRev PW6U satellite checks in from orbit after Transporter 15 launchhttps://www.spacedaily.com/reports/SatRev_PW6U_satellite_checks_in_from_orbit_after_Transporter_15_launch_999.htmlBerlin, Germany (SPX) Nov 28, 2025 SatRev reports that its PW6U Earth observation nanosatellite has completed initial commissioning steps after launch on SpaceX's Transporter 15 rideshare mission, with the first link from orbit confirming that power, communications and basic onboard functions are operating as intended. PW6U is a 6U-class platform built in Wroclaw that extends SatRev's smallsat line into a new generation ofDec 2, 2025 11:26 AM
spacedaily.com
Galileo satellites to launch on Ariane 6 to enhance navigation systemhttps://www.spacedaily.com/reports/Galileo_satellites_to_launch_on_Ariane_6_to_enhance_navigation_system_999.htmlBerlin, Germany (SPX) Nov 21, 2025 On December 17, 2025, two Galileo First Generation satellites, SAT 33 and SAT 34, will be placed in orbit by Ariane 6, launching from Europe's Spaceport in French Guiana. This mission, the fourteenth under the Galileo programme, aims to strengthen the constellation's precision, availability, and operational robustness. Galileo supports billions of users worldwide through navigation functions embNov 22, 2025 7:57 AM
news.ycombinator.com
A Portfolio Ended My 4 Month Job Search and Started My Side Incomehttps://news.ycombinator.com/item?id=45821059Job hunting as a junior AI/ML engineer is brutal. I was about to graduate and spent four straight months applying everywhere, startups, labs, even freelance gigs. Nothing. Not a single callback. At one point, I even looked into those career consulting agencies that promised guaranteed interviews, until I saw the prices. Out of desperation, I figured maybe I could at least make a small portfolio site to look more legit. So I used MGX, nothing fancy. I uploaded my resume PDF and gave it one simple prompt: "Turn this resume into a clean, functional portfolio site deployable on Vercel. Use only the exact information and links from my resume, ensure every link works and all functions run properly. Here are my project GitHub links to include. Please replace the phone number field with a 'Schedule a Call' button linking to my calendar, and keep the section headers clickable for navigation." Two minutes later, I had a fully built page with all my resume sections intact, just a free personal blNov 5, 2025 9:27 AM
news.ycombinator.com
Show HN: Beelzebub (OSS) – MCP "canary tools" for AI agentshttps://news.ycombinator.com/item?id=45158776We’re open-sourcing a simple way to add “canary tools” to AI agents via MCP honeypots. These are functions your agent should never call during normal operation. If a canary is invoked, you get a high-fidelity signal of prompt-injection, tool hijacking, or lateralization—no heuristics, no extra model calls.What it is:- Go framework exposing decoy tools over MCP that look legitimate (names/params/descriptions), return safe dummy output, and emit telemetry when invoked.- Runs alongside your real tools; ship events to stdout/webhook or your pipeline (Prometheus/Grafana, ELK).Why it helps:Agent logs show what happened; canaries mark what must not happen. A single tripwire is an immediate, low-noise indicator of compromise.Real-world relevance (Nx attack):Recent reporting on the Nx npm supply-chain incident (“s1ngularity”) shows malicious versions exfiltrated SSH keys, tokens, and other secrets—and notably abused AI developer tools like Claude/Gemini in the workflow, one of the first documenSep 7, 2025 3:00 PM
nasa.gov
Space Station Cell Studieshttps://www.nasa.gov/missions/station/iss-research/space-station-cell-studies/Science in Space August 2025 Cells are the basic building blocks of all living things, from single-celled bacteria to plants and animals containing vast numbers of them. Cells have adapted for a wide variety of settings and functions. Nerve cells in humans and animals, for example, have long, thin extensions that rapidly transmit signals, while […]Aug 11, 2025 7:00 PM
nasa.gov
NASA Tests Scalable Satellite Tech to Launch Sensors Quickerhttps://www.nasa.gov/centers-and-facilities/langley/nasa-tests-scalable-satellite-tech-to-launch-sensors-quicker/NASA’s Athena Economical Payload Integration Cost mission, or Athena EPIC, is a test launch for an innovative, scalable space vehicle design to support future missions. The small satellite platform is engineered to share resources among the payloads onboard by managing routine functions so the individual payloads don’t have to. This technology results in lower costs […]Jul 18, 2025 10:18 PM
kuhnelo.com
Show HN: Kuhnelo – An almost all-in-one Research Toolkithttps://kuhnelo.com/searchHello HN! Wanted to show an ‘almost’ all-in-one researcher’s toolkit that me and my buddy built up on the side as I was completing my PhD. I call it ‘almost’ because this toolkit was not intended to replace everything a researcher does, but rather it brings together some of the more disparate pieces of software and tools used in research. It was previously released to colleagues and friends as ProjectGIRAF (it’s an inside joke), and we’d love to get HN thoughts on it. Logins/signups are only needed if you want to use the library and search history functions, and it’s using Google sign-on through Kinde, so you don’t have to worry about another account to remember.What’s in Kuhnelo:1. A graph-and-vector search engine – the base database is pretty much OpenAlex, with search being parsed semantically. Currently results can be presented in terms of citations or recency, but we are testing a ‘relevance’ ordering based on our graph centrality measures.You can also click on the author labels tJul 2, 2025 2:28 PM
