I Built an AI Receptionist for a Luxury Mechanic Shop – Part 1

11 mooreds 2026-03-23 10:33:43 UTC itsthatlady.dev Llama Share 6

The user built an AI receptionist named Axle for their brother's luxury mechanic shop to answer customer calls and provide accurate information about prices, hours, and policies. The system uses Retrieval-Augmented Generation (RAG) and a custom-built knowledge base to ensure accurate and grounded responses, and can also collect callbacks for unknown questions.

POSSE – Publish on your Own Site, Syndicate Elsewhere

154 tosh 2026-03-23 08:53:43 UTC indieweb.org Llama Share 33

POSSE (Publish Own Site, Syndicate Everywhere) is a strategy for publishing content on your own site and then syndicating it to other platforms, allowing you to maintain ownership and control over your content while reaching a wider audience. By using POSSE, you can reduce dependence on third-party services, improve search engine rankings, and provide a clear link back to the original ...

PC Gamer recommends RSS readers in a 37mb article that just keeps downloading

634 JumpCrisscross 2026-03-22 18:53:43 UTC stuartbreckenridge.net Llama Share 295

The PC Gamer article has a cluttered interface with multiple ads and popups, including a 37MB initial load. The website continues to download new ads, totaling almost half a gigabyte in five minutes.

Walmart: ChatGPT checkout converted 3x worse than website

86 speckx 2026-03-23 06:53:43 UTC searchengineland.com Llama Share 65

Walmart ended its Instant Checkout deal with OpenAI due to low conversion rates. It will now use its own chatbot, Sparky, to offer a seamless shopping experience within ChatGPT.

Tin Can, a 'landline' for kids

146 tejohnso 2026-03-23 01:53:43 UTC businessinsider.com Llama Share 109

Alison Bennett, a mom in Los Angeles, bought a Tin Can phone for her 8-year-old daughter to chat with friends without a mobile phone. The phone runs off WiFi and has parental controls, allowing kids to communicate with approved contacts during approved hours.

The gold standard of optimization: A look under the hood of RollerCoaster Tycoon

396 mariuz 2026-03-22 19:53:43 UTC larstofus.com Llama Share 115

The author appeared on a German gaming podcast, Stay Forever, to discuss the technology behind RollerCoaster Tycoon. The game's creator, Chris Sawyer, wrote it in Assembly, a low-level language that allowed for high performance.

Can you get root with only a cigarette lighter? (2024)

84 HeliumHydride 2026-03-23 04:53:43 UTC da.vidbuchanan.co.uk Llama Share 16

The user created a low-budget electromagnetic fault injection (EMFI) tool using a piezo-electric lighter and an antenna wire to induce memory errors in a laptop's DDR bus, then used this tool to write a local-privilege-escalation exploit for CPython. The user then used a similar EMFI tool to create a reliable exploit for Linux, allowing them to gain root access by corrupting page tables and ...

Show HN: The King Wen Permutation: [52, 10, 2]

17 gezhengwen 2026-03-23 08:53:43 UTC gzw1987-bit.github.io Full Share 10

The 64 hexagrams of the I Ching have two canonical orderings: the binary natural order (0–63) and the King Wen sequence (~1000 BCE). Treating the map between them as a permutation in S₆₄, its cycle decomposition yields [52, 10, 2] with zero fixed points. 81% of hexagrams are locked in a single cycle. This structure has not been previously reported in the literature.

The future of version control

540 c17r 2026-03-22 15:53:43 UTC bramcohen.com Llama Share 295

Manyana is a project that presents a coherent vision for the future of version control using CRDTs, which provide eventual consistency and informative conflict markers. It offers a better design for version control, including permanent line ordering, informative conflicts, and preserved history, making it a promising proof for CRDT-based version control.

Reports of code's death are greatly exaggerated

418 stevekrouse 2026-03-22 16:53:43 UTC stevekrouse.com Llama Share 310

The user argues that while AI helps with precision in coding, it also creates an illusion of precision, which can lead to bugs and complexity issues when scaled up. The user believes that good code is not just about producing working software, but also about creating beautiful, well-structured abstractions that help master complexity, and that AI will be a boon for coding, not a replacement for it.

The way CTRL-C in Postgres CLI cancels queries is incredibly hack-y

70 andrenotgiant 2026-03-23 04:53:43 UTC neon.com Llama Share 11

When you press Ctrl-C in psql to cancel a query, the client sends an unencrypted CancelRequest message to the Postgres server, which can be intercepted and used to cancel other queries. This is a security risk, but it can be mitigated by using Postgres 18 or later, specifying min_protocol_version=3.2, or using a VPN.

Why I love NixOS

331 birkey 2026-03-22 17:53:43 UTC birkey.co Llama Share 234

The user loves NixOS for its deterministic and reproducible package manager, allowing them to define and rebuild their operating system from a single source of truth. This approach makes experimentation cheap and safe, and the user appreciates its stability, predictability, and coherence across platforms.

Show HN: Agent Kernel – Three Markdown files that make any AI agent stateful

17 obilgic 2026-03-23 07:53:43 UTC github.com/oguzbilgic Llama Share 6

The agent uses three markdown files and a git repo to remember and build on past work without a framework or database. It reads the kernel, updates state and narrative, and adapts to new identities and knowledge.

Project Nomad – Knowledge That Never Goes Offline

463 jensgk 2026-03-22 12:53:43 UTC projectnomad.us Llama Share 155

Project NOMAD is a free, open source offline server that runs on any computer, providing access to knowledge, AI, and maps without internet. It's a self-hosted solution for digital independence, offering a complete library and AI assistant for learning, navigation, and more.

Dataframe 1.0.0.0

17 internet_points 2026-03-23 09:53:43 UTC discourse.haskell.org Llama Share 0

Polars library has reached v1 after 2 years of work, featuring a typed API, Apache Arrow integration, and connectors for large data lakes. The library is now fast and supports data exploration, querying, and integration with AI agents for type-guided data analysis.

Flash-MoE: Running a 397B Parameter Model on a Laptop

358 mft_ 2026-03-22 11:53:43 UTC github.com/danveloper Llama Share 115

Researchers built a 397 billion parameter AI model in 24 hours using C, Objective-C, and Metal shaders on a MacBook Pro. The model runs at 4.4+ tokens/second with production-quality output, using a custom Metal compute pipeline and 48GB RAM.

GoGoGrandparent (YC S16) is hiring Back end Engineers

0 2026-03-23 03:53:43 UTC ycombinator.com Llama Share

GoGoGrandparent is a digital platform helping older adults stay independent at home by adapting on-demand APIs into a concierge-style experience. The company is hiring engineers to join its fully remote team and make a meaningful impact on people's lives.

You are not your job

197 jryio 2026-03-22 15:53:43 UTC jry.io Llama Share 215

Jacob, a software consultant, discusses how people's identities are tied to their jobs, but this is changing with automation and AI, making it essential to focus on relationships and presence rather than technical skills. He encourages readers to reevaluate their sense of self and find value in being present and connecting with others, rather than just their job titles.

A Copy-Paste Bug That Broke PSpice AES-256 Encryption

54 jtsylve 2026-03-23 03:53:43 UTC jtsylve.blog Llama Share 11

SpiceCrypt is a Python library and CLI tool that decrypts encrypted SPICE model files, restoring interoperability between simulators. It exploits a bug in PSpice's encryption scheme, allowing it to recover user keys in seconds and decrypt files that were previously locked to a single simulator.

They’re vibe-coding spam now

97 raybb 2026-03-22 22:53:43 UTC tedium.co Llama Share 55

Spam emails are becoming increasingly difficult to detect due to advancements in AI-powered design, making it harder to tell what's real and what's fake. To protect yourself, use techniques like email obfuscation and aliases, and look for tells such as unusual from addresses and references to your email address rather than your name.

GrapheneOS will remain usable by anyone without requiring personal information

426 nothrowaways 2026-03-22 21:53:43 UTC grapheneos.social Llama Share 120

GrapheneOS remains available internationally without requiring personal information or an account, and it will continue to support Pixel 6a devices until at least July 2027. GrapheneOS has a long-term partnership with Motorola to support their future devices, improving their security to meet GrapheneOS requirements.

MAUI Is Coming to Linux

218 DeathArrow 2026-03-22 15:53:43 UTC avaloniaui.net Llama Share 128

Avalonia MAUI is now available for .NET MAUI, allowing deployment to new platforms like Linux and WebAssembly. This new backend advances Avalonia's vision of cross-platform consistency.

The LCA problem revisited [pdf]

5 remywang 2026-03-23 08:53:43 UTC www3.cs.stonybrook.edu Full Share 1

Five years of running a systems reading group at Microsoft

169 Foe 2026-03-22 17:53:43 UTC armaansood.com Llama Share 51

The user started a reading group at Microsoft in 2021 focused on database internals, which evolved into the Microsoft Systems Reading Group covering various computer science topics, and has been successful due to consistency, organic growth, and collaboration. The group's format, which includes guided reading series and a relaxed atmosphere, has led to valuable connections and learning ...

Ordered dithering with arbitrary or irregular colour palettes (2023)

46 surprisetalk 2026-03-23 00:53:43 UTC matejlou.blog Llama Share 3

The user discusses various methods of image dithering, including ordered dithering and error-diffusion dithering, and their applications to colour palettes of arbitrary or irregular colour distributions. The user also explores more advanced techniques such as N-candidate dithering, barycentric interpolation, and Delaunay triangulation for improving the quality of dithered images.

What young workers are doing to AI-proof themselves

141 wallflower 2026-03-22 18:53:43 UTC wsj.com Full Share 230

Please enable JS and disable any ad blocker

Migrating the American express payment network, twice

76 madflojo 2026-03-23 00:53:43 UTC americanexpress.io Llama Share 28

American Express migrated its payments network twice with zero customer-impacting downtime. The migrations involved transitioning live card authorization traffic to a new, modernized platform.

Building an FPGA 3dfx Voodoo with Modern RTL Tools

195 fayalalebrun 2026-03-22 13:53:43 UTC noquiche.fyi Llama Share 43

The user successfully implemented an FPGA reimplementation of the Voodoo 1 graphics card using SpinalHDL, which allowed them to describe, simulate, and debug the complex design. The user used netlist-aware waveform queries in conetrace to identify and fix a bug that was caused by a stack of small hardware-accuracy mismatches rather than a single catastrophic failure.

LLMs predict my coffee

124 surprisetalk 2026-03-22 20:53:43 UTC dynomight.net Llama Share 50

The user asked for an equation to model the temperature of boiling water in a ceramic coffee mug over time, but found that reality is too complicated and multiple LLMs provided different equations based on exponentially decaying terms. The user then conducted an experiment and found that the actual temperature drop was faster in the first few minutes and slower later on, contradicting the ...

More common mistakes to avoid when creating system architecture diagrams

199 billyp-rva 2026-03-22 11:53:43 UTC ilograph.com Llama Share 59

System diagramming mistakes include poorly labeled resources, disconnected resources, and overly complex diagrams. Common errors also include oversimplification, unnecessary animations, and fan traps, which can lead to confusion and misinterpretation.