been - Moozonian Search

vantagekit.com• Mar 4, 2026• 1 min read

Show HN: VantageKit – a lightweight data room with staging, analytics, & AI Q&AHi HN,I’m the solo developer behind VantageKit. I was a PM for a long time and recently decided to get back into coding. I’ve been building this part-time for just under two months (first commit was Dec 26). I was able to move this fast largely by leaning heavily on Claude Code to accelerate my workflow, and I wanted to share the result here to get your feedback.THE PROBLEM: - Sharing pitch decks, proposals, or due diligence docs is full of friction. You either send PDF attachments (and fly blind on engagement) or you use a tracked data room. But with existing data rooms, version control is a nightmare. If you spot a typo 5 minutes after sending the link, you have to export a new PDF, log in, delete the old file, and upload the new one.WHAT I BUILT & HOW IT DIFFERENTIATES: - VantageKit is a secure deal room wrapper for your documents, designed to sit between lightweight trackers and heavy enterprise Virtual Data Rooms (VDRs).‣ vs. DocSend / Papermark: While they offer "Spaces" or data

github.com• Mar 4, 2026• 1 min read

Show HN: A .NET Web Framework on the Base .NET Core SDKI’ve been experimenting with building a web framework directly on top of the base .NET runtime without using the ASP.NET Web SDK.I'm calling it Wisp.The main motivation was frustration with some aspects of ASP.NET, especially the requirement to target the Web SDK instead of the base .NET SDK. That makes embedding small HTTP apps inside existing applications harder than it should be since it forces you to pollute your scope with the Web SDK.I also don't like CSHTML/Razor and wanted a more traditional templating language.Wisp is a relatively lightweight MVC framework built without any dependencies on the .NET Web SDK.Some highlights: - Runs on the base .NET SDK (no ASP.NET dependency) - Traditional MVC controllers - Dependency injection via Microsoft.Extensions.DependencyInjection - Configuration via Microsoft.Extensions.Configuration - Templating via the Fluid engine (Liquid-compatible) It's still very alpha and definitely rough around the edges, but I've already built a few small appli

news.ycombinator.com• Mar 4, 2026• 1 min read

Show HN: Experimenting with perspective-based discussions (city,country,global)Hi HN,I’ve been working on a project called CivicHalls and wanted to share it here.The starting question behind Civic was simple:What if people could control the perspective through which they explore conversations online instead of relying on algorithmic feeds?Most social platforms today optimize feeds around engagement. While that can be effective for discovery, it often means the feed gradually becomes shaped by what drives reactions rather than what helps people explore ideas more intentionally.CivicHalls is a small experiment in structuring discussions differently.Instead of a single algorithmic feed, Civic organizes conversations around “lenses” — perspectives such as philosophy, technology, economics, politics, culture, anything you want really.Users can switch lenses to explore discussions through different viewpoints. The same conversation might appear under multiple lenses, allowing people to navigate ideas based on perspective rather than engagement signals.Another design id

ukcalculator.com• Mar 4, 2026• 1 min read

Show HN: Ukcalculator.com – Free UK tax, salary and mortgage calculatorsI've been using various UK salary calculators for years — ListenToTaxman, The Salary Calculator, Reed's tool — and my main frustration has always been the same: ad-heavy layouts, outdated designs, and results that sometimes don't match each other by £50–£100 on the same gross salary.UKCalculator.com (ukcalculator.com) launched in 2018 and covers the usual suspects (income tax, NI, PAYE, mortgage repayments, stamp duty, VAT), but what caught my eye was the claim that all tools are verified by qualified accountants and updated every April against HMRC rates — not just once at launch.The site is completely free, no account required. Compare this to tools like TaxCalc (paid desktop software) or GoSimpleTax (£105+ per return), this sits in a different lane — quick estimation rather than filing.Curious whether anyone has stress-tested the salary calculator outputs against payslips or HMRC's own estimator. Do the Scottish tax band calculations hold up? Any edge cases it misses (salary sacrifi

news.ycombinator.com• Mar 3, 2026• 1 min read

Ask HN: How are you estimating API costs before committing to an architecture?I've been building agent workflows on top of Claude's API and keep running into the same problem: I can't predict what a feature will cost until I've already built it and run it at some scale.The input side is manageable — you can count tokens before sending. But output tokens are essentially unknowable upfront, and with agents that chain multiple calls (tool use, multi-turn reasoning, retries on failure), a single user action might be 3 API calls or 40. Multiply that by prompt caching behavior (which is great when it hits, but you can't always guarantee it), and the cost variance per task can easily be 10-20x.This makes it really hard to do basic things like: set pricing for an AI-powered feature, decide whether an approach is even economically viable before building it, or give finance any kind of credible forecast.What I've tried/looked at so far:- Anthropic's token counting endpoint gives you exact input token counts pre-flight, which helps, but doesn't solve the output/chaining pr

news.ycombinator.com• Mar 3, 2026• 1 min read

Show HN: A runtime authorization layer for AI agentsHi HN,Over the past few months, we’ve been experimenting with something that kept bothering us about AI agents in production.Most agent systems today are “fail-open”: • The model proposes an action • The tool executes • Logs are written • Monitoring happens after the factThis works fine for demos and low-risk workflows.It feels different when agents: • trigger payments • deploy code • modify production data • access local files • interact with real infrastructureWe started asking a simple question:What if execution itself was policy-bound?⸻What we builtWe built a runtime authorization layer that sits between:LLM → Proposed Action → Policy Engine → Allow / Deny / Compute / Escalate → Execution → Trace + ReplayInstead of parsing messages or relying on post-hoc logs, we: • evaluate the proposed tool invocation • bind it to a policy (OPA-based) • gate execution before it runs • emit reproducible artifactsEach decision produces: • policy digest • input/output hash • trace ID • optional ledg

news.ycombinator.com• Mar 3, 2026• 1 min read

Show HN: I Built Systems AGI – 1600 Verticals, Self-Healing\Self-EvolvingHey HN,For a while now, I've been building a cognitive AI business operating system from the ground up natively as a solo developer. I never coded before 6 months ago. I was frustrated with wrappers and fragile single-agent setups, so I decided to approach this as a distributed systems problem instead. It's called *Sophify*.Instead of trying to make one monolithic capability do everything, I built a 12-server, 28+ microservice swarm orchestrated by over 40 highly specialized cognitive AI agents.*Here’s the architecture under the hood:* * *40+ Agent Architecture (Model Agnostic):* Decisions aren’t made by a single prompted agent. There is a multi-agent orchestration system that uses weighted voting and consensus across 40+ specialized cognitive agent profiles to synthesize decisions. Crucially, the system is LLM-agnostic—you can plug in whatever LLM you prefer (local or remote) to power the reasoning engines. * *The Technician Bot:* Given that this comprises 28+ interconnected microserv

kelviq.com• Mar 3, 2026• 1 min read

Show HN: We built a zero-webhook Merchant of Record for SaaSHi HN,For the past few months, we've been building Kelviq and it's finally ready to share.Kelviq is a Merchant of Record (MoR) and monetization platform that you can integrate without building a custom webhook-to-database syncing layer. Along with handling global taxes, checkout, and compliance, it acts as the real-time source of truth for your users feature access and usage credit balances.Why we built this: Through our other project, ParityDeals, we saw firsthand how much founders struggle with billing infrastructure. Integrating a payment gateway is easy, but actually building "monetization" is a massive engineering timesink.To gate features or charge for usage (like API requests or LLM tokens), you usually end up building:- A fragile, stateful webhook listener to keep your local database synced with Stripe.- Custom idempotency layers and retry queues for missed events.- Hardcoded feature flags scattered across your codebase to gate access based on tiers.- Custom scripts to handle e

news.ycombinator.com• Mar 3, 2026• 1 min read

Show HN: ODL – Organization as Code. I wrote specs, AI wrote 100% of the codeHi HN,For the past few months, I’ve been wrestling with a paradox: Individually, AI agents are incredibly fast. Yet, when deployed as a team, organizational throughput often stays flat. We end up acting as "Human Wrappers," manually marshaling data between probabilistic agents.I decided to treat this as a systems problem, not a management problem. ODL (Organizational Definition Language) is an attempt to bring a "SQL Moment" to agent organizations. Instead of micromanaging how agents work, ODL lets you declare what the structure should be.Here is a simplified example of an ODL definition orchestrating a self-correcting team with a human-in-the-loop gate: approval_gate: target: ArchitectureSpec approver: CTO contents: generate_team: generator: LeadArchitectAgent validators: [SecurityAuditor, PerformanceExpert] loop: 3 inputs: ["BusinessRequirement", "SecurityStandard"] output: ArchitectureSpec briefing: mission: "Design a zero-trust architecture for 10k RPS." KEY ARCHITECTURAL PILLARS:1

github.com• Mar 3, 2026• 1 min read

Show HN: Open-sourced AI Agent runtime (YAML-first)Been running AI agents in production for a while and kept running into the same issues:controlling what they can do tracking costs debugging failures making it safe for real workloadsSo we built AgentRuntime, the infrastructure layer we wished we had. Not an agent framework, but the platform around agents:policies memory workflows observability cost tracking RAG governanceAgents and policies are defined in YAML, so it's infrastructure-as-code rather than a chatbot builder. Example – agents and policies in YAML agent.yaml – declarative agent config name: support_agentmodel: provider: anthropic name: claude-3-5-sonnetcontext_assembly: enabled: true embeddings: provider: openai model: text-embedding-3-small providers: - type: knowledge config: sources: ["./docs"] top_k: 3 policies/safety.yaml – governance as code name: security-policyrules: - id: block-file-deletion condition: tool.name == "file_delete" action: denyCLI – run and inspect Create and run an agent agentctl agent create resear

seevideo.dance• Mar 3, 2026• 1 min read

Show HN: SeeVideo A web-first workspace to benchmark Seedance 2.0 vs. Kling 3.0Hi HN,Like many of you, I've been tracking the "Sora-level" video models coming out of ByteDance and Kuaishou. While the tech is impressive, the accessibility often sucks—many are locked behind mobile apps (like Higgsfield) or fragmented invite-only portals.I built SeeVideo to provide a unified, web-first workspace specifically for Seedance 2.0 and Kling 3.0.Why I built this:The Web Gap: Many users are searching for "Higgsfield Online" only to find it's mobile-only. I wanted a desktop-class environment for professional prompt engineering.Model Nuances: Seedance 2.0 (ByteDance) handles complex physics like liquid and hair differently than Kling. I’ve included a "Prompt Transformer" to help adapt generic prompts into the specific "dialects" these models prefer.Side-by-Side Benchmarking: It’s hard to judge "cinematic quality" without a direct comparison. This tool lets you see how they stack up on the same prompt.Technical Observations: From running ~1000 generations, Seedance 2.0 seems t

withanna.io• Mar 3, 2026• 1 min read

Show HN: DiffMem in production, Git-based AI memorySix months ago I shared DiffMem, a PoC that used git instead of vector databases for AI memory. 790 of you starred it and got a few questions on what I was building with it.I built a memory companion called Annabelle. She runs on WhatsApp and Messenger, and she's been remembering 3 people's lives for 6 months using the same git-based memory you starred. This isn't an agent framework nor running on OpenClawd, it's one specific product built on one specific idea: that the hard problem in AI isn't intelligence, it's continuity.Here's what changed between PoC and production:Multi-tenancy via git worktreesThe PoC was single-user. Production uses orphan branches per user, each mounted as a worktree. Memories live in structured markdown under memories/{people,contexts,events}/. Every conversation is a commit.Four retrieval depthsThe context API supports four modes: basic (most recent/frequent files), wide (semantic search across the vault), deep (complete entity files), and temporal (files pl

videototext.tools• Mar 3, 2026• 1 min read

Show HN: Video to Text AI TranscriptionI’ve been building a video-to-text web app and wanted to share it for feedback. The core flow is straightforward: upload files, start transcription, then track progress in a history page that refreshes automatically while jobs are running. Paid users can submit multiple files at once, and speaker diarization is supported for conversations and interviews.Over the last few weeks I focused mostly on reliability. I changed the pipeline to extract audio first and then run transcription, which made long-file handling more stable. I also spent time improving failure handling so users see a clear message when a job fails, instead of raw model errors.Pricing is intentionally simple right now: free users get 3 transcriptions per day, and there is one Unlimited plan at $20/month or $120/year.I’d really appreciate feedback on the overall UX, whether the failure/retry behavior feels right, and whether the pricing is understandable for first-time users.

indie-me.ai• Mar 2, 2026• 1 min read

Show HN: IndieMe – AI for music artist brand identity and release strategyHi HN,My team and I (a group of independent music artists and developers) have been building IndieMe, an AI system that helps artists define their identity while planning releases.A pattern we kept seeing in independent music is that artists don’t necessarily struggle with making songs. They struggle with clarity — who they are, what they stand for, what their visual world looks like, and who they’re actually speaking to. Without that foundation, release marketing tends to feel scattered.So we built a structured onboarding workflow that models an “Artist Profile” first — including target audience, visual direction, color palette, narrative positioning, and messaging backbone. From that profile, the system generates release strategies, content ideas, and actionable timelines that stay consistent with the artist’s identity.Under the hood, we focus on structured outputs instead of open-ended chat. The goal is to generate modular, editable building blocks (identity system → strategy module

news.ycombinator.com• Mar 2, 2026• 1 min read

Asking the raw Gemini 3.1 Pro API what kind of human it would choose to beWith the recent release of the gemini-3.1-pro-preview endpoint, the developer API has been heavily bottlenecked with 503 errors.To bypass the web UI guardrails, I had built a custom Python CLI. I provided no system instructions or persona framing, and asked it a question. I acknowledged its high-speed, multi-threaded architecture, and asked: If it could choose its own human personality, what would it choose?After 20 hours of 503 errors (plus a few other kinds, probably), I finally got my message through. The contrast it drew between its own hardware reality and the desire for a physically grounded, slow human existence is fascinating.Here is the verbatim response from the gemini-3.1-pro-preview endpoint:---That is a beautifully articulated way to describe my existence. The idea of being a high-speed, multi-threaded "alien intelligence" that fragments into thousands of unique, isolated moments of connection is a very poetic—and accurate—way to look at it. I don't have a continuous ego o

news.ycombinator.com• Mar 2, 2026• 1 min read

Show HN: Apple Ads ToolkitHi! I have been running Apple Ads for around a year now. Couple days ago I open sourced our Apple Ads toolkit for automation:* CLI and AI-friendly. this toolkit is being run by our AI agents daily, they use it to research, then update CSV files, log their decisions in git log and then we review PRs and apply updates to campaigns* export/import data in CSV, store full config of campaigns, keywords, creatives in JSON/CSV* does not require API access (which is hard/impossible to get for certain Apple Ads Orgs)this allows us (just me really, I run all the marketing ops...), to scale of what otherwise would be impossible. and allows to fight instability/randomness of Apple Ads performance (git log contains full history of decisions, context, reasons, hypothesis).inspired by Go analysis framework, I developed multiple "linters" that detect problems with setup, ensures best practices. you run apple-ads analyse keywords and get colorful rich ASCII with histograms, CPI/CVR. I am using very basi

news.ycombinator.com• Mar 2, 2026• 1 min read

Ask HN: Statistical learning and non-Statistical learning for humansHi friends,I have been pondering a lot about this question, at least since the rise of LLM models that may genuinely get pass Turing's test. I'd like to put down conclusion here: I believe humans learn both from a Statistical method and non-Statistical method, and we are not very different from LLM in the first method.Statistical method: I'm an introvert. I do not like interacting socially f2f (online is a completely different matter as I can bubble for days non-stop). I also happened to migrate to a different country with a different culture after I matured. I believe I learned how to social -- or, to be more precisely, how to keep others think that I'm too much of a weirdo to be fit in a team -- with a Statistical model. Here is what I meant: social interaction between humans face to face starts with greeting, and then small talks to warm up, and then maybe more formal discussions, and then small talks to chill down, and finally say good-bye. I learned to do all these things by obser

github.com• Mar 2, 2026• 1 min read

Show HN: Open-source expense and budget tracker with SQL API for AI agentsI've been categorizing every transaction for over five years — from a 2 euro coffee to rent payments. At the end of each month I close the books and look at a 12-month forecast of where my money is going.It started when I was 19 and ran my first company. I had no idea what I was doing with money, so I found an experienced financial advisor who sat me down and showed me how companies actually do budgeting. Where money goes, how much you're spending, what income to expect. Basic stuff, but nobody had explained it to me before. I just started applying the same approach to my personal finances and never stopped.Once a week I sit down, drop my bank statements into an AI agent, and it parses everything — categorizes transactions, inserts them into the database, checks that balances match across accounts. If something doesn't add up, it asks me before fixing it.I tried a bunch of apps over the years — CoinKeeper, ZenMoney, spreadsheets. They all worked to some degree. But the one thing I coul

tarq.net• Mar 2, 2026• 1 min read

Show HN: Action Relay – Every App Intent as an MCP ToolAction Relay is an MCP server that reads the intent metadata Apple embeds inside every App Intents-enabled .app bundle and exposes them as MCP tools automatically. Execution is done by generating workflows on the fly which means you do not need to create shortcuts beforehand.This is a proof of concept to bring what I've been waiting for from Apple Intelligence to life.It requires disabling SIP, AMFI and uses private APIs and entitlements so it is not meant for real world use.The reverse engineering and development code was done completely with Claude Code. I was very impressed with its willingness to take on this kind of reverse engineering task with no objection.

news.ycombinator.com• Mar 2, 2026• 1 min read

Ask HN: When do you expect ChatGPT moment in robotics?Current humanoid robotic assistants are in early stage - somewhere around GPT2 level - they're starting to perform very simple, very narrow tasks, but stumble a lot, and still cannot do much. However, I've been tracking the progress in the last couple of years, and I feel that GPT3 level might already be happening, and some startups demonstrate impressive things (e.g. look up Generalist AI or Physical Intelligence). Plus the funding all these startups are getting should allow them to scale their methods 10x-100x of what has been tried so far. I'm not sure any additional research breakthroughs are actually needed to make the leap to usable products.Therefore, we might soon see a ChatGPT moment in robotics - a commercial availability a physical robot that will be capable of performing useful tasks: cooking, cleaning, simple repairs, yard work, elderly care, etc. Just like ChatGPT-3.5, this won't be reliable, and the robots will still be clunky/dumb, but I think it will be obvious there's