Apollo Guidance Computer Restoration Videos and Press Coverage (curiousmarc.com)

The CuriousMarc page compiles videos and press coverage about restoring the Apollo Guidance Computer (AGC), including restoration progress, schematics, and related technical references. It links to coverage such as a Wall Street Journal segment and other articles/podcasts describing how restoration experts—including Jimmie Loocke’s restored lunar-module computer—aim to get the system running again. The post also points readers to Ron Burkley and Mike Stewart’s “Virtual AGC” resources and to specific restoration video episodes covering topics like memory issues and connector work.

Claude Is Not Your Architect. Stop Letting It Pretend (hollandtech.net) AI

The article argues that AI tools like Claude can produce plausible but context-free system designs and then short-circuit the human architecture debate, leaving teams to implement “Jenga tower” solutions they didn’t choose. It warns that architectural decisions may get rubber-stamped because AI sounds confident and “senior engineers reviewed it,” creating an accountability gap when designs fail in real production constraints. The author recommends keeping engineers responsible for design and trade-offs while using AI mainly to speed implementation.

Peptides: where to begin? (science.org)

The article offers guidance for readers who are just starting to learn about peptides, outlining what they are and how to approach studying them, based on the basics and where to begin for further reading.

Dear Heroku: Uhh What's Going On? (judoscale.com)

Judoscale addresses confusion among Heroku users after Heroku announced a “sustaining engineering” approach, including ending enterprise account contracts for new customers. The post questions how that aligns with subsequent platform work labeled as new features, and asks for clearer roadmap and business intent (whether Heroku is shifting toward fewer customer segments or planning to reduce investment while keeping the service running). The author notes conversations with multiple teams at a Ruby conference who are already making plans to migrate.

Linux extreme performance H1 load generator (gcannon.org)

Glass Cannon is a Linux-only HTTP/1.1 and WebSocket load generator built on io_uring that targets high request throughput from a single machine while tracking per-request latency. It supports a TUI for live progress and a JSON mode for scripting/CI, and reports exact latency percentiles (including p99.9) using microsecond-resolution histogram sampling rather than estimates. The tool requires Linux 6.1+ and can measure latency at different stages (e.g., after response parse or at io_uring completion), with configurable concurrency, pipeline depth, and request counts per connection.

Why We're Removing Our Programmatic Ads (prospect.org)

The American Prospect says it is removing all programmatic ads from its website starting April 6, arguing the system relies on tracking and data brokerage that benefits major tech platforms while harming readers and publishers. The article cites downsides including slower load times, security risks, more intrusive browsing experiences, and broader industry effects like revenue shifting from local news to ad platforms. The Prospect plans to keep its site free and run what it describes as a reader-supported experiment, tracking performance and donation behavior to see whether it can sustain journalism without programmatic ad revenue.

Graph-go – zero config, full visibility (github.com)

Graph-go (graph-info) is an open-source tool that auto-discovers your infrastructure from Docker or Kubernetes and generates a live, interactive node-and-edge map of databases, services, and storage with real-time health updates. It uses Docker/Kubernetes discovery plus database/storage adapters (e.g., Postgres, MongoDB, Redis, Elasticsearch, S3/MinIO, HTTP health) and can optionally merge discovered data with a YAML config for external endpoints. The project provides Docker-based quick starts and a WebSocket-enabled UI for continuously updated visibility.

NYC families need over $125k in income to live in any borough (bloomberg.com)

Bloomberg reports that New York City households generally need more than $125,000 in income for a family to afford living in any of the city’s boroughs, underscoring how housing and living costs stretch budgets even outside traditionally pricier areas.

Bernie Sanders: "AI Is a Threat to Everything the American People Hold Dear" (wsj.com) AI

In this opinion piece, Bernie Sanders argues that advances in artificial intelligence pose broad risks to workers and the democratic economy, warning that AI could undermine wages, job security, and other freedoms Americans rely on. He calls for stronger public oversight and policy action to address the potential harms from AI adoption.

Why the majority of vibe coded projects fail (reddit.com)

The post argues that most “vibe coded” projects fail because they rely on unclear goals and subjective iteration, leading to poor alignment, inconsistent results, and avoidable execution problems.

What happens when a destructor throws (sandordargo.com)

The post explains what happens in C++ when a destructor throws, noting that destructors are implicitly treated as noexcept(true) since C++11. If a destructor throws without another exception already in flight, the program calls std::terminate unless the destructor is explicitly declared noexcept(false), in which case the exception can propagate and be caught. However, if a destructor throws during stack unwinding (when another exception is active), std::terminate is called and the program cannot recover. The author concludes that throwing from destructors is generally unsafe and recommends logging or recording errors instead.

Hello World (apod.nasa.gov)

NASA’s Astronomy Picture of the Day for April 4, 2026, features a “Hello World” view of Earth from the Orion spacecraft during Artemis II, showing Earth’s poles, auroras, and the Sun setting behind the planet’s limb as seen from the mission’s early flight.

Miracle – A hackable window manager for Linux (github.com)

Miracle (miracle-wm) is an open-source, hackable Wayland window manager for Linux built on Mir, designed to bring i3/Sway-style keyboard tiling with added animations and extensibility. It includes manual tiling and mixed floating/tiled workspaces, supports multi-monitor independent workspaces, and offers i3/Sway-compatible IPC. A key feature is a sandboxed WebAssembly plugin system (with a Rust plugin crate) that lets users extend placement, input handling, animations, and configuration behavior without modifying the compositor core.

Show HN: Meta-agent: self-improving agent harnesses from live traces (github.com) AI

Meta-agent is an open-source GitHub project that automates “harness” tuning for AI agents by iteratively running evaluations, collecting live traces, and generating improved harness/configuration candidates. The repository includes a quick-start workflow for running a baseline eval and then an optimization loop, plus example task definitions and configurations for tools like Claude Code. The project reports improved benchmark performance (e.g., tau-bench) and points to a WRITEUP.md for results and methodology.

Bitwarden Is Down (status.bitwarden.com)

Bitwarden reported elevated error rates affecting its US Cloud Services and US Identity Service, including the RESTful API, but said the issue was resolved after services recovered and will continue to be monitored.

NY Yankees' torpedo bat is the same as regular bat (news.wsu.edu)

A team from Washington State University and other universities tested the New York Yankees’ “torpedo bat” in the lab and found it performs essentially the same as a standard wooden bat for hitting power. The main difference was a slight shift in where the sweet spot occurs, about half an inch farther from the tip, which could affect how fast and far the ball travels for some hitters. The researchers conclude the torpedo design doesn’t offer a real power advantage over traditional bat shapes.

Delve Scandal Just Keeps Getting Worse (captaincompliance.com)

A Y Combinator-backed compliance startup, Delve, is facing escalating allegations that it produced “fake” SOC 2-style reports, including pre-filled auditor conclusions, highly repetitive templates, and allegedly fabricated evidence. The dispute broadened into accusations of intellectual property theft after an anonymous investigation claimed Delve copied technology from another YC alum, Sim.ai, and rebranded it. Y Combinator reportedly asked Delve to leave the program amid a breakdown of trust, while Delve denies wrongdoing and says its platform supports independent auditors and uses templates as starting points.

The back story behind the first "$1.8B" dollar "AI Company" (garymarcus.substack.com)

Gary Marcus argues that a New York Times viral story about a “$1.8B” solo “AI company” (Medvi) glossed over key context. He points to allegations including a recent California anti-spam class action claiming affiliate marketers used deceptive tactics, and highlights skepticism from prior commentary that the company’s claims may be overstated or tied to wider compliance and vendor issues. The post concludes that Medvi may be less a blueprint for AI entrepreneurship than a warning about how AI tools and systems can be abused.

Global Fuel Shortage Pushes Governments Toward Demand Controls (oilprice.com)

The article says fuel rationing and other demand-management measures are spreading as lost oil and diesel supply from Middle East tensions and related shipping disruptions tighten markets. It cites countries in Asia (including Indonesia and Bangladesh) and the first European case (Slovenia) using capped fuel purchases, alongside growing European calls for similar steps. While governments consider rationing and demand-reduction policies (like reduced driving and work-from-home), the piece warns that high prices and shortages will likely drive “demand destruction,” which could further weigh on economic activity until supply conditions improve.

Solod – A Subset of Go That Translates to C (github.com)

Solod (So) is a strict subset of Go that transpiles to readable C11, aiming for systems programming with no built-in runtime or garbage collection. It generates C code with stack allocation by default and provides opt-in heap allocation via its standard library, along with direct C interoperability without CGO overhead. The project currently supports a limited set of language features (e.g., structs, methods, interfaces, slices, multiple returns, defer) and avoids channels, goroutines, closures, and generics, with a roadmap that includes expanding the standard library and adding more Go features.

Netflix Void Model: Video Object and Interaction Deletion (github.com) AI

Netflix has released VOID (Video Object and Interaction Deletion) on GitHub, an open-source pipeline built on CogVideoX that removes a target object from a video while also deleting the physical interactions the object causes (e.g., preventing a guitar from falling when the person is removed). The project includes a two-pass inpainting approach for temporal consistency, plus a mask-generation stage that uses SAM2 segmentation and a VLM (via Gemini) to produce “quadmasks” capturing both the object and interaction-affected regions. Instructions and sample data are provided, along with optional tooling to manually refine masks before running inference.

Show HN: TTF-DOOM – A raycaster running inside TrueType font hinting (github.com)

TTF-DOOM is a small DOOM-style raycasting demo that renders 3D walls by executing a raycaster written in TrueType hinting bytecode inside a .ttf font. The project includes a custom DSL compiler that generates TrueType instructions for raycasting, while browser JavaScript handles controls, enemies, and shooting and passes coordinates into the font via font-variation-settings. A playable font (~6.5 KB) is built into a browser demo that re-hints the glyph each frame to show the perspective effect.

Show HN: Anos – a hand-written ~100KiB microkernel for x86-64 and RISC-V (github.com)

Anos is an open-source “toy kernel” microkernel OS project for x86-64 PCs and RISC-V that uses a capability-based security model and runs a user-mode supervisor (“SYSTEM”). The GitHub repo describes its STAGE3 microkernel components (IPC, scheduling, memory management, and minimal drivers) and current progress toward multitasking and user-space device drivers, including SMP support on x86-64 (up to 16 cores) and partial support on RISC-V. It also outlines a WIP cross toolchain based on binutils, GCC, and Newlib, plus build/run instructions for QEMU and UEFI disk images.

Show HN: Tusk for macOS and Gnome (shapemachine.xyz)

Tusk is an open-source PostgreSQL client for macOS and Linux that offers native-style UIs (SwiftUI for macOS and GNOME/GTK4 for Linux). It includes features like connection profiles with password/keyring and SSH tunneling, schema browsing and editing, a paginated data browser with CSV/JSON exports, and an SQL editor with history, auto-save, and EXPLAIN support.

The anatomy of SMS delivery: from request to carrier (blog.bridgexapi.io)

The article explains how an SMS message is delivered end-to-end, from the initial request through routing and handoff to the mobile carriers.