![Telli (YC F24) Is Hiring Eng, Design, Growth [On-Site, Berlin]](/cdn-cgi/image/width=1200,quality=75,format=auto,fit=cover/https://careers.telli.com/image/attachment%3A5928c484-9890-4c7b-848a-357e73bcced6%3ASocialMediaPreviewImage.png?table=block&id=8af41504-5c69-45da-89d8-c8a1a40a0b9c&spaceId=f7a4b93e-acd7-45c8-bf0e-bef97019e9b7&width=2000&userId=&cache=v2)
Telli (YC F24) Is Hiring Eng, Design, Growth [On-Site, Berlin]
Telli, a Y Combinator‑backed startup, has built the first AI‑native phone system and reached over $2 million in annualized revenue with a 12‑person team. The platform lets companies automate repetitive calls with AI voice agents while managing human agents on the same interface. To accelerate growth, Telli is hiring engineers, designers, and growth professionals in Berlin, offering referral bonuses up to €10,000. The company aims to replace low‑quality outsourced call centers and improve customer experience at scale.
Noora Health (YC W14) Is Hiring AI/ML Engineer
Noora Health, the Y Combinator‑backed health‑tech startup, announced a full‑time AI/ML Engineer position in Bengaluru with a salary of ₹2.5M‑₹3.5M INR. The role involves building end‑to‑end AI systems to enhance its caregiver education platform, which has already trained over 30 million...
The Five Levels: From Spicy Autocomplete to the Dark Factory
Dan Shapiro outlines a five‑tier model for AI‑driven software development, borrowing the NHTSA’s automotive automation levels. He argues that cheap AI compute is driving a rapid deflation of code costs, forcing teams to shift from manual coding to AI‑generated output....
TrueVault (YC W14) Is Hiring a Growth Lead to Test Different Growth Channels
TrueVault, a Y Combinator‑backed privacy‑focused SaaS for SMBs, is hiring its first Growth Lead. The remote, US‑based role will design and test multiple acquisition channels, own the end‑to‑end sales motion, and drive a 15‑25% quarterly increase in qualified pipeline. Compensation...
Unrolling the Codex Agent Loop
OpenAI’s Codex CLI introduces an agent loop that orchestrates user prompts, model inference, and tool executions to generate reliable software changes. The loop builds a role‑ordered prompt, streams model responses via SSE, and feeds tool outputs back into subsequent queries....
European Alternatives
European alternatives for digital services are gaining traction as businesses seek compliant, locally‑based SaaS and cloud solutions. The initiative highlights advantages such as GDPR‑aligned data protection, streamlined VAT handling, and uniform EU legal frameworks. By choosing European providers, companies can...
Ghostty's AI Policy
Ghostty has introduced a strict AI usage policy that mandates full disclosure of any AI tools used in contributions, limits AI‑generated pull requests to accepted issues, and requires human verification of all AI‑assisted code. The policy bans AI‑generated media, enforces...
Vibe a Guitar Pedal
Polyend’s Endless is a $299 guitar effects pedal that lets users generate custom sounds by describing them in plain language or coding them. Its beta Playground converts text prompts into playable patches within minutes, and the device runs a single...
Scaling PostgreSQL to Power 800M ChatGPT Users
OpenAI scaled a single‑primary Azure PostgreSQL flexible server to serve 800 million ChatGPT users, leveraging nearly 50 read‑replicas across global regions. By combining aggressive query tuning, connection pooling with PgBouncer, and off‑loading write‑heavy workloads to sharded Cosmos DB, the platform now handles...
Capital One to Acquire Brex for $5.15B
Capital One announced a $5.15 billion cash‑and‑stock acquisition of fintech firm Brex, slated to close mid‑2026. The deal gives Capital One access to Brex’s corporate‑card platform and expense‑management software used by firms like DoorDash and Robinhood, extending its presence to over...
AnswerThis (YC F25) Is Hiring
AnswerThis, a Y Combinator‑backed AI agent orchestration startup, has reached $1 million ARR in just eight months and serves over 200,000 daily users from institutions like Stanford and MIT. The company is building a system of record for scientists that integrates...
Show HN: Dotenv Mask Editor: No More Embarrassing Screen Leaks of Your .env
The Dotenv Mask Editor is a Visual Studio Code extension that provides a grid‑based interface for editing .env files while automatically masking any value longer than six characters. Masked entries appear as asterisks in the view and only reveal their...
Lix – Universal Version Control System for Binary Files
Lix is a universal version‑control system that can diff binary and structured files such as PDFs, Excel sheets, and JSON, delivering human‑readable, field‑level changes. It layers on top of any SQL database, leveraging ACID guarantees while exposing versioned tables via...
Show HN: Rails UI
Rails UI, a new UI component library for Ruby on Rails, has moved out of beta and is now live with a suite of free and premium themes. It offers plug‑and‑play forms, buttons, layouts, and full‑app designs that integrate seamlessly...
Waiting for Dawn in Search: Search Index, Google Rulings and Impact on Kagi
A U.S. court ruled in August 2024 that Google is a monopoly in general search services, and the Department of Justice’s 2025 remedies require Google to provide its web index and search results to qualified competitors on non‑discriminatory terms. The ruling...
TrustTunnel: AdGuard VPN Protocol Goes Open-Source
AdGuard has open‑sourced its proprietary VPN protocol, now called TrustTunnel. The protocol blends VPN traffic with HTTPS using TLS and HTTP/2/3, aiming to evade detection and throttling. TrustTunnel’s code and specification are released under a permissive license, with client apps...
Three Types of LLM Workloads and How to Serve Them
The article classifies large‑language‑model (LLM) inference into three workload categories—offline, online, and semi‑online—and explains how each demands distinct engineering trade‑offs. Offline jobs focus on batch throughput, online tasks require sub‑hundred‑millisecond latency, while semi‑online workloads need flexible autoscaling for bursty demand....
I Scanned 2,500 Hugging Face Models for Malware/Issues. Here Is the Data
Veritensor launches as a zero‑trust security solution for the AI model supply chain, offering deep static analysis and cryptographic verification of popular model formats such as Pickle, PyTorch, Keras, GGUF, and Python wheels. The tool can detect malicious code—including RCE,...
Hightouch (YC S19) Is Hiring
Hightouch, the Series C‑backed data activation platform, announced a broad hiring push covering customer success, engineering, marketing, partnerships, people, and sales roles. The company highlights its remote‑first model with offices in San Francisco, NYC, Charlotte and London, and offers competitive compensation, equity,...
Nukeproof: Manifesto for European Data Sovereignty
NukeProof Alliance, launched by Finnish storage firm SpaceTime, aims to create a sovereign European cloud by uniting local providers, startups, MSPs and telcos. The initiative responds to the dominance of US hyperscalers—Google, Microsoft and Amazon—who now control over 60% of...
The Challenges of Soft Delete
Soft delete is common, but using an "archived_at" flag often bloats tables, complicates queries, and slows maintenance. The article outlines why retained rows rarely get read and how they inflate backups. It then evaluates four PostgreSQL‑centric alternatives: application‑level archiving to...
Our Approach to Age Prediction
OpenAI is rolling out an age‑prediction model for ChatGPT consumer plans to identify accounts likely belonging to users under 18. The system evaluates behavioral and account‑level signals and automatically applies stricter content safeguards for minors. Users mistakenly flagged can verify...
Show HN: Fence – Sandbox CLI Commands with Network/Filesystem Restrictions
Fence is an open‑source, container‑free sandbox that isolates CLI commands by default, blocking all outbound network traffic and enforcing configurable filesystem permissions. It lets developers define allow‑list domains, writable paths, and prohibited commands via a simple JSON file, making it...
Dockerhub for Skill.md
The Official SKILLS.md Hub launches on Dockerhub as a centralized registry for SKILLS.md files that extend AI assistants such as Claude and ChatGPT. Users can search and install skills via the simple "sr" CLI, accessing tools ranging from Slack automation...
QMD - Quick Markdown Search
QMD is an on‑device hybrid search engine that indexes markdown notes, meeting transcripts, and documentation, offering BM25 keyword, vector semantic, and LLM‑re‑ranked queries—all executed locally via node‑llama‑cpp with GGUF models. The tool provides a CLI for collection management, embedding generation,...
KAOS – The Kubernetes Agent Orchestration System
KAOS (Kubernetes Agent Orchestration System) is an open‑source, Kubernetes‑native framework that lets developers deploy, manage, and orchestrate AI agents at scale. It introduces Agentic Graphs to represent distributed agent networks as native resources and uses the Model Context Protocol for...
Legal Structures for Latin American Startups (2021)
Latin American founders often mis‑choose legal structures, risking massive tax liabilities—as illustrated by a startup that paid over $100 M to the U.S. Treasury after using a Delaware C‑Corp. The guide, based on 80+ investments by Magma Partners, outlines when to...
Apple Testing New App Store Design that Blurs the Line Between Ads and Results
Apple is testing a new App Store search layout on iOS 26.3 that removes the blue background traditionally used to highlight sponsored results. The only visual cue left is a small “Ad” label next to the app icon, making ads blend...
Show HN: Pipenet – A Modern Alternative to Localtunnel
Pipenet is a modern, self‑hostable alternative to Localtunnel, offering bundled client and server packages for exposing local services to the internet. It supports custom subdomains, multiple domains, TypeScript, ES modules, and single‑port cloud deployment. Users can connect to a public...
Show HN: Kacet – a Freelancer Marketplace with Crypto-Native Payments
Kacet launches a crypto‑native freelance marketplace that connects blockchain developers, auditors, AI builders, and other web3 talent with clients seeking on‑chain expertise. The platform charges a flat 5 % fee and uses Solana smart‑contract escrow to lock payments until work is...
Show HN: AskUCP – UCP Protocol Explorer Showing All Products on Shopify
AskUCP, launched on January 11 2026, is a global AI shopping search built on the Universal Commerce Protocol (UCP). UCP, an open standard co‑created by Google and Shopify, lets applications query product data across e‑commerce platforms without traditional APIs or integrations. At...
Show HN: Lume 0.2 – Build and Run macOS VMs with Unattended Setup
Lume 0.2 is an open‑source VM runtime that runs macOS and Linux on Apple Silicon using Apple’s Virtualization Framework, delivering near‑native performance. It ships as a single binary offering both a CLI and an HTTP API for headless VM creation, automated...
Show HN: Show HN: I Built a Fail-Closed Watchdog for ESP32/Arduino in C++
Qrystal has released official SDKs that let ESP32 and other IoT devices report health data to its Qrystal Uplink monitoring service. The kits include Arduino‑C++, native ESP‑IDF C++, and MicroPython libraries, each offering lightweight heartbeat APIs, TLS encryption, connection pooling,...
Show HN: Figma-Use – CLI to Control Figma for AI Agents
The open‑source figma-use CLI lets developers and AI agents control Figma directly from the terminal, bypassing the heavyweight MCP JSON protocol. By accepting JSX, the tool leverages LLMs’ native React knowledge to generate frames, text, components, and auto‑layout with minimal...
Show HN: I Built a Tool to Assist AI Agents to Know when a PR Is Good to Go
Good To Go (GTG) is a new command‑line tool that deterministically assesses whether a GitHub pull request is ready to merge. It aggregates CI statuses, classifies review comments, and tracks thread resolution, returning a single, unambiguous status. The tool outputs...
Show HN: Streaming Gigabyte Medical Images From S3 without Downloading Them
WSI Streamer is a cloud‑native tile server that serves whole‑slide medical images directly from S3 without downloading the full file. By issuing HTTP range requests, it extracts only the bytes required for each JPEG tile and delivers them instantly through...
Ask HN: Is It Still Worth Pursuing a Software Startup?
The Ask HN thread debates whether software startups still have a defensible moat, noting that large firms rarely copy products instantly due to internal bureaucracy. Commenters argue that speed, focus, brand credibility, and user trust form real competitive advantages. Many startups...
Zep AI (Agent Context Engineering, YC W24) Is Hiring Forward Deployed Engineers
Zep AI, a 2023 YC W24 startup specializing in agent context engineering, announced multiple senior hires including Staff Engineer, Lead Forward Deployed Engineer, and Developer Relations Lead. Its platform stitches together chat history, business data, and user behavior to deliver...
6-Day and IP Address Certificates Are Generally Available
Let’s Encrypt has made six‑day short‑lived certificates and IP address certificates generally available. The short‑lived certificates expire after 160 hours, reducing the exposure window when a private key is compromised. IP address certificates, supporting both IPv4 and IPv6, are required to...
The Spectrum of Isolation: From Bare Metal to WebAssembly
The article maps the spectrum of execution environments from bare‑metal servers through virtual machines, containers, process sandboxes, and language‑specific virtual environments, ending with emerging models like serverless and WebAssembly. It explains how each layer isolates code, resources, and dependencies, highlighting...
Show HN: Hc: An Agentless, Multi-Tenant Shell History Sink
HC (History Collector) is an open‑source, agentless service that streams shell command history from any server into a centralized PostgreSQL store. It supports plain TCP and TLS ingestion, multi‑tenant isolation via API‑key authentication, and provides a plain‑text, grep‑friendly export endpoint....
Building a Better Bugbot
Cursor’s Bugbot, an AI‑driven code‑review agent, has evolved from a prototype to an agentic architecture that now scans over two million pull requests each month. By introducing a custom resolution‑rate metric and running 40 systematic experiments, its bug‑resolution rate climbed from...
Show HN: Aventos – An Experiment in Cheap AI SEO
Aventos introduced an AI search analytics platform aimed at small businesses, offering a forever‑free tier and paid plans from $10 to $23 per month. The service monitors brand mentions in AI‑driven search results such as ChatGPT and Google AI Overviews,...
Aviator (YC S21) Is Hiring to Build Multiplayer AI Coding Platform
Aviator, a Y Combinator S21 graduate, is hiring senior and full‑stack engineers to expand its AI‑powered software development platform. The company’s suite—including MergeQueue, FlexReview, and the newly launched Runbooks—automates code merges, review routing, and collaborative workflow automation. Its tools already...
Show HN: Tusk Drift – Turn Production Traffic Into API Tests
Tusk Drift CLI is an open‑source record‑and‑replay system that turns production API traffic into deterministic test suites. It ships with Node.js and Python SDKs and integrates with Tusk Drift Cloud for trace storage, regression classification, and CI/CD enforcement. An AI‑powered setup wizard can automatically...
Show HN: Control Claude Permissions Using Cloud Hooks
Rulebricks released Claude Code Guardrails, a lightweight plug‑in that intercepts Claude’s tool calls via a PreToolUse cloud hook and routes decisions to the Rulebricks API. The system lets teams publish, edit, and version guardrail rules in a web UI, with...
Selling SaaS in Japan
Japan represents a massive SaaS opportunity, but its sales process diverges sharply from Western markets. Prospects first download product documentation, then engage in a prolonged, consensus‑driven evaluation before any demo. This slower cadence yields higher retention but demands localized content,...
We Gave Our Browser Agent a 3MB Data Warehouse
The 100x bot now embeds PGLite, a Postgres engine compiled to WebAssembly, giving the browser agent a 3 MB relational data warehouse for long‑term memory. This lets users store scraped data in tables, run fuzzy joins, hierarchical ltree queries, and window‑function...
Nao Labs (Open-Source Analytics Agent, YC X25) Is Hiring
nao Labs, a Y Combinator Spring 2025 startup, is hiring a founding software engineer to build its open‑source AI analytics agent. The company’s AI‑powered IDE already serves over 100 data teams, and the new agent will enable chat‑based analytics across...
Show HN: MailPilot – Freedom to Go Anywhere While Your Agents Work
MailPilot is a terminal‑based user interface that wraps local large language model agents and forwards each interaction to an email thread. The tool lets users and their AI agents communicate asynchronously via email, freeing the user from a constant desk...