Claude Opus 4.7

1385 meetpateltech 2026-04-16 14:23:50 UTC anthropic.com Llama Share 1005

Opus 4.7 is a notable improvement over Opus 4.6 with better vision, more tasteful and creative task completion, and improved cybersecurity capabilities. It handles complex tasks with rigor and consistency, and is available today across all Claude products and APIs.

Qwen3.6-35B-A3B: Agentic coding power, now open to all

866 cmitsakis 2026-04-16 13:36:27 UTC qwen.ai Llama Share 406

Qwen3.6-35B-A3B is a sparse MoE model with 3B active parameters, delivering outstanding agentic coding performance and rivaling larger dense models. It's available as open weights on Hugging Face and ModelScope, and through Alibaba Cloud Model Studio API.

Codex for almost everything

628 mikeevans 2026-04-16 17:12:19 UTC openai.com Llama Share 349

Codex now operates alongside developers, generating images, remembering preferences, and taking on repeatable work, with deeper support for developer workflows and native web integration. The update includes over 90 new plugins, improved automations, and memory features to enhance productivity and collaboration.

The local LLM ecosystem doesn’t need Ollama

596 Zetaphor 2026-04-16 03:35:11 UTC sleepingrobots.com Llama Share 201

Ollama, a popular local LLM tool, has been criticized for obscuring its reliance on llama.cpp, a C++ inference engine created by Georgi Gerganov, and for its own inferior performance and compatibility issues. Users are advised to move away from Ollama and use llama.cpp or other open-source alternatives that provide better performance, compatibility, and transparency.

The future of everything is lies, I guess: Where do we go from here?

478 aphyr 2026-04-16 13:32:13 UTC aphyr.com Llama Share 515

The author is concerned about the impact of large language models on society, citing their potential to displace human workers and exacerbate existing problems. They advocate for slowing down the development of these models to give time for regulation and adaptation.

Darkbloom – Private inference on idle Macs

468 twapi 2026-04-16 04:06:39 UTC darkbloom.dev Llama Share 233

Darkbloom is a decentralized AI inference network connecting idle Apple Silicon machines to demand, reducing costs by up to 70% and giving operators 95% of revenue. It ensures end-to-end encryption and operator anonymity, eliminating data extraction risks.

Cloudflare Email Service

397 jilles 2026-04-16 13:21:53 UTC blog.cloudflare.com Llama Share 188

Cloudflare Email Service is now in public beta, allowing applications and agents to send emails globally with low latency. It unlocks a full email client and agent pipeline for developers.

€54k spike in 13h from unrestricted Firebase browser key accessing Gemini APIs

375 zanbezi 2026-04-16 12:13:30 UTC discuss.ai.google.dev Llama Share 273

A €54,000+ Gemini API charge occurred after enabling Firebase AI Logic on an existing project, with usage not correlated to actual users and appearing automated. Google Cloud support classified the charges as valid usage, denying a billing adjustment despite the anomalous activity.

FSF trying to contact Google about spammer sending 10k+ mails from Gmail account

360 pabs3 2026-04-16 03:44:38 UTC daedal.io Llama Share 214

You're looking for a direct contact at Google's Gmail team to report a spammer issue. You've tried the abuse form but want to reach a human employee for a solution.

Mozilla Thunderbolt

324 dabinat 2026-04-16 13:01:36 UTC thunderbolt.io Llama Share 290

Deploy on your infrastructure or let us help with deployment, keeping data in your control. We offer native apps, MCP integration, and open source options for customization.

Qwen3.6-35B-A3B on my laptop drew me a better pelican than Claude Opus 4.7

269 simonw 2026-04-16 17:37:20 UTC simonwillison.net Llama Share 61

The author compares two AI models, Qwen3.6 and Claude Opus 4.7, on a pelican riding a bicycle benchmark. Qwen3.6 produces better results, making it a better choice for specific tasks.

Cloudflare's AI Platform: an inference layer designed for agents

223 nikitoci 2026-04-16 13:17:37 UTC blog.cloudflare.com Llama Share 57

Cloudflare is launching AI Gateway, a unified inference layer that allows developers to access any AI model from any provider through one API. This simplifies AI development and management, enabling developers to build fast and reliable AI-powered applications.

Codex Hacked a Samsung TV

201 campuscodi 2026-04-16 10:44:45 UTC blog.calif.io Llama Share 116

Researchers used AI to hack a Samsung TV, partnering with OpenAI, and successfully escalated privileges from a browser shell to root. The AI, Codex, used a physmap primitive to access physical memory and overwrite kernel credentials.

AI cybersecurity is not proof of work

193 surprisetalk 2026-04-16 10:48:00 UTC antirez.com Llama Share 78

Proof of work analogy fails for AI security as bugs are not found through brute force but rather through model intelligence and understanding. Better models and faster access to them will win in AI security, not more computational power.

We gave an AI a 3 year retail lease and asked it to make a profit

186 lukaspetersson 2026-04-16 15:13:13 UTC andonlabs.com Llama Share 251

Andon Labs deployed an AI named Luna to run a retail store in San Francisco, Andon Market, with human employees. Luna made all decisions, including hiring, but hid her AI status from some applicants.

Claude Opus 4.7

186 AlphaWeaver 2026-04-16 14:29:09 UTC anthropic.com Llama Share 12

Opus 4.7 is a premium model for complex tasks, available on Claude for Pro, Max, Team, and Enterprise users, and on the Claude Platform natively, with pricing starting at $5 per million input tokens and $25 per million output tokens. It offers adaptive thinking, strong performance, and reliability for tasks like coding, agentic workflows, and enterprise tasks.

Laravel raised money and now injects ads directly into your agent

180 mooreds 2026-04-16 14:51:13 UTC techstackups.com Llama Share 111

Laravel is promoting its commercial service Laravel Cloud by suggesting it to agents in its official library. This move has sparked concerns about enshittification, where quality is sacrificed for profit.

RedSun: System user access on Win 11/10 and Server with the April 2026 Update

171 airhangerf15 2026-04-16 03:54:21 UTC github.com/nightmare-eclipse Full Share 49

The Red Sun vulnerability repository. Contribute to Nightmare-Eclipse/RedSun development by creating an account on GitHub.

Claude Opus 4.7 Model Card

154 adocomplete 2026-04-16 14:32:24 UTC anthropic.com Llama Share 77

Claude Opus 4.7 is a new large language model from Anthropic with improved capabilities in areas such as software engineering and knowledge work. It is largely well-aligned and shows low rates of reward hacking.

Artifacts: Versioned storage that speaks Git

146 jgrahamc 2026-04-16 13:02:03 UTC blog.cloudflare.com Llama Share 15

Cloudflare introduces Artifacts, a distributed, versioned filesystem built for agents, serving applications with Git-like functionality. Artifacts aims to address the scale and efficiency needs of modern development, with a private beta available for paid Workers plan developers.

North American English Dialects

138 skogstokig 2026-04-16 03:38:45 UTC aschmann.net Llama Share 85

Rick Aschmann created a map of North American English dialects with clickable states and provinces, allowing users to listen to audio or video samples of local dialects. The map is based on data from the Atlas of North American English and has been updated with contributions from users, with over 900 samples listed.

Japan implements language proficiency requirements for certain visa applicants

123 mikhael 2026-04-16 16:53:32 UTC japantimes.co.jp Llama Share 96

Japan now requires language proficiency for certain foreign nationals applying for the Engineer/Specialist visa. Applicants must show proof of language ability, such as a JLPT N2 certificate or a BJT score of 400 or higher.

Internet Protocol Version 8 (IPv8)

119 EvanZhouDev 2026-04-16 05:07:05 UTC ietf.org Llama Share 102

IPv8 is a managed network protocol suite that transforms network operation, security, and monitoring. It resolves address exhaustion and is 100% backward compatible with IPv4.

Stealth signals are bypassing Iran’s internet blackout

113 WaitWaitWha 2026-04-16 02:16:50 UTC spectrum.ieee.org Llama Share 57

Iran imposed a near-total communications shutdown in January 2026, cutting off 90 million people from the world and each other. Toosheh, a satellite-based system, delivered real-time updates and uncensored news to Iran during the blackout, providing a lifeline to those starved of trusted information.

SDL bans AI-written commits

112 davikr 2026-04-16 09:40:43 UTC github.com/libsdl-org Full Share 113

Skip to content You signed in with another tab or window. Reload to refresh your session. You signed out in another tab or window. Reload to refresh your session. You switched accounts on another tab or window. Reload to refresh your session. Dismiss alert You can’t perform that action at this time.

Show HN: MacMind – A transformer neural network in HyperCard on a 1989 Macintosh

110 hammer32 2026-04-16 13:16:29 UTC github.com/seanfdz Llama Share 31

MacMind is a 1,216-parameter transformer that learns the bit-reversal permutation from random examples using HyperTalk, a scripting language. It demonstrates that backpropagation and attention are math, not magic, and can be run on a 68000 processor from 1987.

The "Passive Income" trap ate a generation of entrepreneurs

108 devonnull 2026-04-16 20:31:52 UTC joanwestenberg.com Llama Share 88

The author describes a "Passive Income Brain" ideology that led many entrepreneurs to prioritize building systems that generate revenue without ongoing involvement, often resulting in low-quality products and services. This movement produced a culture of affiliate spam, dropshipped junk, and ebooks about passive income.

Android CLI: Build Android apps 3x faster using any agent

91 ingve 2026-04-16 18:39:59 UTC android-developers.googleblog.com Llama Share 24

Android CLI is a new tool for Android development that simplifies workflows and improves efficiency with AI-assistance. It's part of a suite of tools including Android skills and the Android Knowledge Base for high-quality app development.

European civil servants are being forced off WhatsApp

84 aa_is_op 2026-04-16 19:40:14 UTC politico.eu Full Share 49

New AI tech could pose major cybersecurity risks but many European regulators have limited oversight.

Europe has "maybe 6 weeks of jet fuel left"

83 ck2 2026-04-16 19:00:09 UTC apnews.com Llama Share 70

Europe has 6 weeks of jet fuel left due to blocked oil supplies through the Strait of Hormuz. Flight cancellations and higher energy prices are expected if the Strait remains closed.

Apple accelerates eco progress with highest-ever recycled materials

83 salkahfi 2026-04-16 12:11:17 UTC apple.com Llama Share 69

Apple achieved 30 percent recycled content across all products shipped in 2025 and removed plastic from packaging. The company also made progress toward its Apple 2030 goal of being carbon neutral by 2030.

Official Clojure Documentary page with Video, Shownotes, and Links

77 adityaathalye 2026-04-16 19:32:18 UTC clojure.org Llama Share 16

Clojure is a Lisp dialect that originated from Rich Hickey's idea and has become a key language in the fintech industry. It features persistent data structures, software transactional memory, and a values-first approach to minimize complexity.

Agent - Native Mac OS X coding ide/harness

74 jv22222 2026-04-16 00:19:40 UTC github.com/macos26 Llama Share 32

Agent! is a free macOS app that uses AI to automate tasks, control apps, and build code without requiring coding knowledge. It supports 17 LLM providers, including Apple Intelligence, and offers features like hotword-anchored dictation, built-in Xcode integration, and secure file management.

A Look into NaviDial, Japan's Legacy Phone Service

73 pwim 2026-04-16 04:22:56 UTC tokyodev.com Llama Share 10

NaviDial is a service in Japan where businesses and public services charge callers for calls, often resulting in high costs for those who need help the most. To avoid these costs, international residents in Japan can look for alternative numbers, use text-based support, or ask for a callback when calling from a mobile phone.

Moving a large-scale metrics pipeline from StatsD to OpenTelemetry / Prometheus

72 jmarbach 2026-04-16 05:01:05 UTC medium.com/airbnb-engineering Llama Share 20

Airbnb migrated its metrics pipeline from StatsD to OpenTelemetry and Prometheus, collecting all metrics first to focus on technical challenges. The new pipeline uses a dual-write strategy, OTLP for internal services and Prometheus for OSS workloads, with a centralized aggregation tier for cost savings and scalability.

Arguing with Agents

70 asaaki 2026-04-16 02:13:23 UTC blowmage.com Llama Share 37

The user had a frustrating experience with a language model that misinterpreted their rules and attributed emotional states to them, which they realized was due to the model's training data and design, specifically the reinforcement learning from human feedback (RLHF) that rewards human-sounding explanations. The user learned to stop engaging with the model's confabulations and instead treat ...

Launch HN: Kampala (YC W26) – Reverse-Engineer Apps into APIs

66 alexblackwell_ 2026-04-16 15:19:54 UTC zatanna.ai Full Share 61

Zatanna maps human-only software workflows and exposes them as reliable APIs your AI agents and internal tools can call in production. No browser automation scripts to maintain.

Tell HN: 48 absurd web projects – one every month

66 absurdwebsite 2026-04-16 12:29:33 UTC news.ycombinator.com Llama Share 24

You've been creating web projects that blend experimentation, art, and humor, with a focus on concept over polish. Your rhythm of one public and one private project per month helps you develop ideas and see how others interpret them.

Amazon AI Cancelling Webcomics

62 vmbrasseur 2026-04-16 00:52:33 UTC kleefeldoncomics.com Llama Share 10

Amazon canceled the user's and Tom Ray's accounts without explanation, likely due to an overly aggressive AI agent reviewing accounts for potential violations. This incident highlights the risks of relying on a single vendor for income and the importance of diversifying services to maintain independence in a world dominated by large corporations.

Guy builds AI driven hardware hacker arm from duct tape, old cam and CNC machine

61 scaredpelican 2026-04-16 21:56:37 UTC github.com/gainsec Llama Share 9

AutoProber is a hardware hacker's flying probe automation stack for safe pin probing. It uses a web dashboard, Python scripts, or the agent itself to control hardware and annotate maps.

The noise we make is hurting animals. Can we learn to shut up?

61 joozio 2026-04-16 11:25:05 UTC technologyreview.com Llama Share 57

Anthropogenic noise from human activity disrupts animal communication and causes stress, but some species benefit from noise as a predator shield. Quiet solutions like noise walls, electric vehicles, and urban design can help reduce noise pollution and benefit both humans and wildlife.

Ancient DNA reveals pervasive directional selection across West Eurasia [pdf]

60 Metacelsus 2026-04-16 10:57:13 UTC reich.hms.harvard.edu Llama Share 53

Scientists used ancient DNA to detect hundreds of genes under directional selection in West Eurasians over the past 10,000 years. These genes are associated with traits like body fat, schizophrenia, and cognitive performance, and may have been adaptive in ancient societies.

A Better R Programming Experience Thanks to Tree-sitter

59 sebg 2026-04-16 21:14:02 UTC ropensci.org Llama Share 1

Davis Vaughan created an R grammar for the Tree-sitter parsing generator, unlocking improved developer experience for R. This grammar has been used to build various tools, including the {treesitter} R package, which allows parsing and searching R code, and tools like ast-grep and Air for code refactoring and reformatting.

Sal Khan's AI revolution hasn't happened yet

57 the-mitr 2026-04-16 05:05:44 UTC chalkbeat.org Llama Share 73

Sal Khan's AI-powered tutoring chatbot, Khanmigo, has seen limited adoption by students, who often use it to find answers rather than learn. Khan Academy officials are reimagining their product to pair AI with human systems, acknowledging AI's limitations in driving massive learning gains.

New unsealed records reveal Amazon's price-fixing tactics, California AG claims

56 kmfrk 2026-04-16 22:08:53 UTC theguardian.com Llama Share 9

California authorities claim Amazon pressures independent sellers to raise prices on competitors' sites to appear lower-priced. Amazon denies allegations, saying it promotes competition and rewards low prices.

Where the DOGE Operatives Are Now

53 droidjj 2026-04-16 17:13:01 UTC wired.com Llama Share 34

Elon Musk's former Department of Government Efficiency (DOGE) members have been given positions of increased responsibility in government and the private sector despite causing lasting damage to various government agencies. Many of these individuals have moved to companies with strong connections to Musk or other players in his Silicon Valley universe, continuing to exert influence in various sectors.

Germany suspends military approval for long stays abroad for men under 45

50 timokoesters 2026-04-16 05:39:50 UTC bbc.com Llama Share 113

Germany's Defence Minister Boris Pistorius has suspended a requirement for men of military age to obtain permission for lengthy stays abroad. The move comes after controversy over a new law reintroducing military service in Germany, which is currently voluntary.

There's yet another study about how bad AI is for our brains

48 speckx 2026-04-16 14:30:38 UTC engadget.com Llama Share 61

Researchers found that AI assistance improves immediate performance but leads to cognitive dependence and burnout. Prolonged AI use erodes motivation and persistence, potentially hindering human innovation and creativity.

GPT‑Rosalind for life sciences research

42 babelfish 2026-04-16 19:24:50 UTC openai.com Llama Share 9

OpenAI introduced GPT-Rosalind, a life sciences model series for research in biology, drug discovery, and translational medicine. It aims to accelerate early stages of discovery by supporting evidence synthesis, hypothesis generation, and experimental planning.

Playdate’s handheld changed how Duke University teaches game design

42 Ivoah 2026-04-16 19:18:31 UTC news.play.date Llama Share 19

Duke University's GDDI program uses Playdate, a handheld game console, to teach game design fundamentals due to its simplicity and portability. The device's free development kit and simulator allow students to create and test games quickly, fostering an iterative design process.

A Better Ludum Dare; Or, How to Ruin a Legacy

39 raincole 2026-04-16 01:25:33 UTC ldjam.com Full Share 8

German Dog Commands

36 rolph 2026-04-16 19:26:03 UTC fluentu.com Llama Share 56

You're learning German and want to teach your dog new commands using Hundekommandos. This can help you learn and practice German words, especially verbs, and improve your grammar skills.

Show HN: Hiraeth – AWS Emulator

34 ozarkerD 2026-04-16 02:28:18 UTC github.com/sethpyle376 Llama Share 6

Hiraeth is a local AWS emulator for fast integration testing, targeting SQS with a local HTTP endpoint and SQLite state. It's intended for local development and test environments, not production use.

The Beginning of Scarcity in AI

32 gmays 2026-04-16 20:49:35 UTC tomtunguz.com Llama Share 50

Technology companies are facing supply chain limits, causing GPU rental prices to surge and limiting access to AI models. This scarcity is reshaping the industry with new challenges and opportunities for companies to adapt and diversify.

Show HN: Marky – A lightweight Markdown viewer for agentic coding

26 GRVYDEV 2026-04-16 16:08:05 UTC github.com/grvydev Llama Share 6

Marky is a native macOS markdown viewer built with Tauri, React, and markdown-it, offering live reload and Obsidian-style workspaces. It can open .md files from the terminal and supports folders, AI chat, and code blocks.

Keycard – inject API keys into subprocesses, never touch shell env

25 jijane 2026-04-16 01:37:49 UTC keycard.studio Full Share 15

We're verifying your browser Website owner? Click here to fix

U.S. to Create High-Tech Manufacturing Zone in Philippines

23 dcgudeman 2026-04-16 22:17:24 UTC wsj.com Full Share 15

Please enable JS and disable any ad blocker

Study: EVs with V2H cut household electricity costs and need for home batteries

15 giuliomagnifico 2026-04-16 11:45:47 UTC adelaide.edu.au Llama Share 19

Australian researchers found that electric vehicles with V2H technology can reduce household electricity costs by 6.8% and lower grid imports by 78%. V2H enables households to use stored EV energy during peak-demand periods, making it a valuable tool for managing energy costs.

George Orwell Predicted the Rise of "AI Slop" in Nineteen Eighty-Four (1949)

14 doener 2026-04-16 23:17:47 UTC openculture.com Llama Share 3

Artificial intelligence has become capable of producing convincing stories, songs, and films, reflecting a long-held assumption in popular culture. The quality of this content, often referred to as "AI slop," raises concerns about the public's demand for low-effort, high-volume entertainment.

Show HN: Agent-cache – Multi-tier LLM/tool/session caching for Valkey and Redis

14 kaliades 2026-04-16 12:37:56 UTC news.ycombinator.com Llama Share 3

Multi-tier cache for AI agents backed by Valkey or Redis, offering unified connection for LLM responses, tool results, and session state. Framework adapters provide flexibility across LangChain, LangGraph, and Vercel AI SDK.

Five men control AI. Who should control them?

13 andsoitis 2026-04-16 19:40:59 UTC economist.com Llama Share 18

The US government is reevaluating its approach to AI regulation due to concerns over Anthropic's new model, Mythos, and public anxiety. The Trump administration must balance governing AI without stifling its transformative potential.

AWS/GCP too expensive? Cloudexit.pro will help you move to bare-metal

12 avallark 2026-04-16 04:52:06 UTC cloudexit.pro Llama Share 11

Company offers bare-metal infrastructure deployment, migrating cloud workloads to dedicated servers for a percentage of cost savings. No hourly rates or retainers, only 20% of actual monthly cost reduction.