The hottest Technology Substack posts right now

And their main takeaways
Category
Top Technology Topics

35 Thoughts About AGI and 1 About GPT-5

Am I Stronger Yet? • 3855 implied HN points • 14 Aug 25
đź•ą Technology Artificial Intelligence Machine Learning Software Development Data science Human-computer interaction
  1. Current AI can't really match human intelligence. Even though it can do some complex tasks, there are still many things it struggles with, like understanding context or learning continuously.
  2. Humans can learn new skills from just a few examples, while AI often needs a lot of data to learn. This difference is why humans pick up things like driving so much faster than AI systems.
  3. As AI technology advances, it may start playing a bigger role in complex tasks. This could change how we work and interact with machines, possibly making us more like spectators in our own jobs.

Weaponized Inventory: How the U.S.–China Tech War Broke the Supply Chain

Gad’s Newsletter • 38 implied HN points • 09 Mar 26
đź•ą Technology Semiconductors Supply Chain AI hardware Robotics Trade policy Manufacturing Procurement
  1. Sudden changes in export rules are triggering massive over-orders for AI chips that overwhelm testing, licensing, and shipping systems, so companies must add regulatory scenario planning to their demand forecasts.
  2. Most rare-earth refining and midstream processing are concentrated and slow to replicate, creating hidden Tier‑N chokepoints that require deep BOM traceability and years of investment to resolve.
  3. Complex products like humanoid robots hinge on a few hard-to-replace precision parts and long supplier‑qualification timelines, forcing a costly shift from just-in-time sourcing to resilience-focused, multi-source supply networks.

Containerized Missile Systems Without Cranes

Tapa’s Substack • 4 HN points • 05 Oct 24
đź•ą Technology Defense Logistics Engineering Aerospace Innovation
  1. Containerized missile systems aim to fit missiles into shipping containers for easy transport. This could help with quick deployment and keeping them hidden.
  2. Most missiles are too tall for standard shipping containers, requiring them to be laid down horizontally. This makes launching them more complicated.
  3. A new idea suggests using a small jump jet to lift and angle the missile for firing, making it faster and potentially cheaper than using a crane system.

On Evaluating Cognitive Capabilities in Machines (and Other "Alien" Intelligences)

AI: A Guide for Thinking Humans • 462 implied HN points • 14 Jan 26
đź•ą Technology Artificial Intelligence Machine Learning Cognitive Science Benchmarks Interpretability
  1. Benchmarks can be misleading: high scores don’t prove real-world understanding because models can rely on training leaks, shortcuts, or narrow task-specific tricks.
  2. Evaluation should borrow rigorous methods from developmental and animal cognition: avoid anthropomorphic assumptions, run control and adversarial experiments, and test robustness with novel variations to see if abilities truly generalize.
  3. Go beyond accuracy to study mechanisms and failures: distinguish competence from performance, analyze error types, and publish negative or replication results to understand what models really do.
Get a weekly roundup of the best Substack posts, by hacker news affinity:

Four Things: What Do These People Know That We Don't?

State of the Future • 29 implied HN points • 27 Feb 26
đź•ą Technology AI Compute Hardware Software Startups Security Workforce
  1. AI builders expect rapid, widespread disruption of white‑collar work, so societies will need to adapt fast to avoid big economic and employment shocks.
  2. The next big gains will come from orchestration, not just bigger chips or models — combining diverse hardware and specialised components will be a key competitive edge.
  3. Models and models' outputs are now attackable and competitive assets, so security and new architectures (many small agents checking each other) are becoming essential to reduce errors and theft.

Elon Musk’s terrifying vision for AI

Marcus on AI • 10908 implied HN points • 16 Feb 25
đź•ą Technology Artificial Intelligence Machine Learning Data Privacy Tech Ethics Big Tech
  1. Elon Musk's AI, Grok, is seen as a powerful tool for propaganda. It can influence people's thoughts and attitudes without them even realizing it.
  2. The technology behind Grok often produces unreliable results, raising concerns about its effectiveness in important areas like government and education.
  3. There is a worry that Musk's use of biased and unreliable AI could have serious consequences for society, as it might spread misinformation widely.

The Shape of Artificial Intelligence

The Algorithmic Bridge • 806 implied HN points • 22 Dec 25
đź•ą Technology Artificial Intelligence Machine Learning AI Ethics AI safety Technology Policy Societal Impact
  1. AI abilities are spiky and alien, with huge strengths in narrow domains and surprising failures on simple, commonsense tasks. This jagged shape means AI won't neatly fill a human-shaped general intelligence anytime soon.
  2. Human intelligence grew slowly through biological evolution while AI is created by mathematical optimization and market pressures, so AIs develop different strengths and can expand much faster in specific directions. This difference produces distinct "Umwelten" and makes AI growth uneven and hard to predict.
  3. The useful approach is practical coexistence: learn the geometry of AI, use it to augment tasks where its spikes help, keep humans in the loop where its valleys remain, and stop assuming full replacement is the default outcome. This mindset favors designing systems that combine human and AI strengths rather than chasing a single notion of AGI.

Grok 3 Beta in Shambles

Marcus on AI • 10750 implied HN points • 19 Feb 25
đź•ą Technology Artificial Intelligence Machine Learning Computing Data science Software Development
  1. The new Grok 3 AI isn't living up to its hype. It initially answers some questions correctly but quickly starts making mistakes.
  2. When tested, Grok 3 struggles with basic facts and leaves out important details, like missing cities in geographical queries.
  3. Even with huge investments in AI, many problems remain unsolved, suggesting that scaling alone isn't the answer to improving AI performance.

Meaningmaking at work.

The Uncertainty Mindset (soon to become tbd) • 259 implied HN points • 21 Aug 24
đź•ą Technology AI Development Product Management Human-Machine Interaction Software Design
  1. AI tools often fail because they can't understand the deeper meaning behind our decisions. They confuse what humans can intuitively interpret.
  2. Meaningmaking is crucial in many business processes. Humans make subjective decisions all the time that machines simply can't replicate.
  3. To create better AI products, we need to separate meaningmaking tasks from other work. This helps us design tools that support human decision-making instead of trying to replace it.

New company, old playbook?

The Engineering Manager • 5 implied HN points • 20 Mar 26
đź•ą Technology Leadership Management Teams Onboarding Strategy Engineering
  1. You arrive as both an expert and a beginner, so hold your experience lightly and adopt a beginner's mind to stay curious and open to how things actually work here.
  2. Use the first 30 days for a listening tour and simple assessments—listen more than you act, resist quick fixes, and learn who and why things are the way they are.
  3. In days 30–90 pick your battles, steer without doing, and land visible results that set the right tone; bring guiding principles with you but leave behind one-size-fits-all prescriptions.

Vibe Coding and the Maker Movement

Technically • 94 implied HN points • 26 Feb 26
đź•ą Technology AI Software Product Open Source Infrastructure Community
  1. Vibe coding skipped the slow, playful "scenius" phase of earlier maker cultures and went straight into production, so people can build fast but often lack the practical judgment that comes from long, messy practice.
  2. Think of vibe coding as consuming a surplus of machine intelligence: spent well it produces taste, attention, reputation, or gift-like social capital, but spent badly it’s just addictive, disposable output.
  3. Long-term value tends to accumulate in the model and infrastructure layers unless creators intentionally capture the byproduct signal as datasets, documentation, or curated taste, and framing the work as consumption can help avoid burnout.

Apple’s $20 Billion Ticking Time Bomb

Big Technology • 5504 implied HN points • 13 Jun 25
đź•ą Technology Big Tech AI Software Internet Antitrust
  1. Apple relies heavily on payments from Google, which are about $20 billion a year. If these payments disappear, Apple's services revenue could significantly drop.
  2. The potential loss of Google's payments is a serious risk for Apple, especially since its services segment is its only growing revenue source right now.
  3. If the court decides to cut Google's payments, Apple may struggle to find a replacement income that matches the profits, which could lead to financial issues for the company.

Deliberative Alignment, And The Spec

Astral Codex Ten • 11149 implied HN points • 12 Feb 25
đź•ą Technology AI Ethics Machine Learning Computer Science Tech governance Human-computer interaction
  1. Deliberative alignment is a new method for teaching AI to think about moral choices before making decisions. It creates better AI by having it reflect on its values and learn from its own reasoning.
  2. The model specification is important because it defines the values that AI should follow. As AI becomes more influential in society, having a clear set of values will become crucial for safety and ethics.
  3. The chain of command for AI may include different possible priorities, such as government authority, company interests, or even moral laws. How this is set will impact how AI behaves and who it ultimately serves.

Mise en place for agentic coding

Dev Interrupted • 98 implied HN points • 19 Feb 26
đź•ą Technology Software Development Artificial Intelligence Developer Tools Task management Product Design DevOps
  1. Spend time on mise en place before coding so agents know exactly what you want; clear preparation (briefing, spec, task breakdown) makes implementation much faster and reduces debugging.
  2. Practice context fluency by encoding domain knowledge, value judgments, and constraints so agents can make aligned micro-decisions without guessing.
  3. Keep the toolchain simple and remove extra layers so your thinking maps directly to execution; simpler interfaces let agents deliver the right architecture quickly.

Turing Award Histories

The Chip Letter • 4149 implied HN points • 26 Jul 25
đź•ą Technology Computer Science Engineering Interviews History Innovation
  1. The Computer History Museum has a treasure trove of almost 2,000 interviews with important figures in computer science, offering insights into the field and its pioneers.
  2. These interviews capture not just technical knowledge but also the personal stories of innovators, making them relatable and engaging for anyone interested in technology.
  3. The Turing Award winners have made significant contributions and their interviews provide a curated starting point for exploring this vast archive of oral histories.

Why Ads on ChatGPT Are More Terrifying Than You Think

The Algorithmic Bridge • 1104 implied HN points • 02 Dec 25
đź•ą Technology AI Advertising Media User Experience Computing
  1. Ads in ChatGPT will change how it gives information, making it less about what the user needs and more about what advertisers want.
  2. The shift to ads means OpenAI's focus will be on making money from advertisers instead of helping users, which could hurt the user experience.
  3. Blending ads into AI responses could lead to more misinformation, as users won't easily recognize when they are being marketed to.

Dear Conservatives: Shape the AI Future Before It's Shaped for You

The Dossier • 121 implied HN points • 13 Feb 26
đź•ą Technology Artificial Intelligence Tech Policy Ethics Regulation Industry Investment Innovation Society
  1. Conservatives should stop treating AI as an enemy and actively engage as entrepreneurs, investors, technologists, and customers to help shape its direction.
  2. If conservatives don’t participate, AI systems will be designed by a narrow tech elite and their philosophical assumptions will get baked into training data, safety rules, and product norms.
  3. The window to influence AI is closing because power and infrastructure are consolidating and regulation will be slow, so act now to insert conservative values into mainstream systems rather than waiting or building isolated alternatives.

Subtypes and status-dependent data: pure relational approach

Minimal Modeling • 304 implied HN points • 29 Jan 26
đź•ą Technology Databases SQL Data Modeling Software Engineering
  1. Lock a subtype/status column to a single value with a CHECK so subtype tables can only hold rows for that exact status, and reference the main table with a composite foreign key (id, status) to prevent contradictory data.
  2. Give the main table a unique (id, status) pair and make subtype tables include a defaulted, immutable status plus their own keys so you can model both single- and multi-row status-specific information without NULLs.
  3. This is a pure relational, NULL-free way to encode subtypes/status-dependent data using only standard constraints (CHECK, PK, FK), moving integrity into the schema and making the design extensible even if it isn’t commonly taught.

The Internet Needs to Be Smashed and Rebuilt

TK News by Matt Taibbi • 10189 implied HN points • 22 Feb 25
đź•ą Technology Internet Media Government Information Empowerment
  1. The Internet has become a barrier to understanding and access to information. It used to help people, but now it's harder to find reliable news.
  2. Many people question the trustworthiness of news sources, indicating a general distrust in media. This makes it tough to locate credible information.
  3. There's a call for a major overhaul of the Internet to make it a better tool for knowledge and empowerment again. The idea is to rebuild it from the ground up.

8 plots that explain the state of open models

Democratizing Automation • 451 implied HN points • 07 Jan 26
đź•ą Technology Artificial Intelligence Machine Learning Open Source AI adoption AI Industry
  1. Chinese open models—especially Qwen—now dominate downloads, finetunes, and general adoption across the ecosystem, often outpacing many other providers combined.
  2. New entrants and recent Western releases show only limited adoption so far, with older Western models like Llama still widely downloaded while GPT-OSS shows early promise but hasn’t shifted overall usage.
  3. The clearest competitive opportunity is at large model scales, where DeepSeek and a few others outperform Qwen’s big models, but Chinese models still lead on benchmarks with only a few competitors getting close.

Reading List 01/18/25

Construction Physics • 11483 implied HN points • 18 Jan 25
đź•ą Technology Space Tech Real Estate Infrastructure Transportation
  1. Real estate development plays a big role in how skyscrapers look and are built. There are great books that explain the process and thinking of developers involved in these projects.
  2. Congestion pricing in New York is improving traffic speeds significantly in a short time. People entering the zone are moving faster, helping them save time and frustration during their commutes.
  3. Some homes in Los Angeles survived wildfires due to smart design choices that included careful landscaping and construction techniques. These details can make a big difference in fire-prone areas.

A $30M plan to take back social media from billionaires

Taylor Lorenz's Newsletter • 12301 implied HN points • 13 Jan 25
đź•ą Technology Social media Decentralization Digital Infrastructure
  1. A new campaign called FreeOurFeeds aims to take social media back from billionaires. They want to make social media a public good for everyone.
  2. The project plans to raise $30 million to build a new social media system that gives users more control and allows for better community interactions.
  3. The goal is to create a decentralized social media environment where users can express themselves freely without corporate or political pressures.

Data Science Weekly - Issue 562

Data Science Weekly Newsletter • 179 implied HN points • 29 Aug 24
đź•ą Technology Data science Artificial Intelligence Machine Learning Data Engineering Statistics
  1. Distributed systems are changing a lot. This affects how we operate and program these systems, making them more secure and easier to manage.
  2. Statistics are really important in everyday life, even if we don't see it. Talks this year aim to inspire students to understand and appreciate statistics better.
  3. Understanding how AI models work internally is a growing field. Many AI systems are complex, and researchers want to learn how they make decisions and produce outputs.

🤖 Humanoid robots arrive just as humans stop showing up

Faster, Please! • 456 implied HN points • 15 Jan 26
đź•ą Technology Robotics Artificial Intelligence Automation Workforce Demographics Economic Growth
  1. The U.S. is heading into demographic decline: deaths are projected to exceed births by 2030 and the total population is expected to stop growing by the mid-2050s and then shrink.
  2. Fewer births and an aging population will squeeze the labor force and threaten economic growth, and without immigration the country would already be getting smaller.
  3. Physical AI and humanoid robots are increasingly seen as a timely solution to fill labor gaps and help keep the economy growing, rather than just as job destroyers.

Is n8n / low-code automation dead?

Enterprise AI Trends • 232 implied HN points • 01 Feb 26
đź•ą Technology AI Automation Low-code Enterprise SaaS Product Design Developer Tools
  1. Natural-language, markdown-first automation tools challenge the assumption that non-technical users need visual drag-and-drop builders, because describing automations in plain English can produce deterministic, scalable workflows for complex AI tasks.
  2. Visual low-code tools are not dead but their role is evolving; enterprises will adopt natural-language automation gradually, leading to hybrid stacks and different tools for different problems.
  3. Product teams, operators, executives, and investors must reevaluate tool choices, training, renewals, and investments because bets on visual workflow platforms may be riskier as natural-language automation gains traction.

Roche Axelios Pricing

ASeq Newsletter • 72 implied HN points • 27 Feb 26
đź•ą Technology Genomics Sequencing Instruments Pricing Market Biotech
  1. Roche’s Axelios is priced competitively with Illumina — offering $150 per duplex genome and very low simplex read costs — but not so cheap that it will immediately displace Illumina, so adoption will be gradual.
  2. Roche has clear advantages over newer rivals: it’s lower risk, more technically interesting, and cheaper for many counting/simplex applications, so it’s likely to outcompete companies like Ultima and Element.
  3. Reusable chips and low per-run chip costs give Roche room to cut prices or offer big customer discounts later, but high switching costs and Illumina’s entrenched position mean market changes will be slow and uneven.

The Future of Economic Security with Dan Kim and Chris Miller

ChinaTalk • 296 implied HN points • 21 Jan 26
đź•ą Technology Semiconductors Industrial Policy Supply Chains National Security Geopolitics Manufacturing Export Controls
  1. A modest CHIPS budget can’t fully de-risk the U.S. from foreign suppliers, so policy should aim for resilience — building key clusters, mature-node capacity, and capability — rather than unaffordable self-sufficiency.
  2. Measure economic security with clear metrics like the Four Cs (capacity, capability, competition, criticality) and practical goals such as minimizing “time to recovery,” while creating institutions and incentives to execute and coordinate industrial strategy.
  3. There’s a trade-off between invention (high-value innovators) and fast-following scale-ups: both matter for national power, and friend-shoring or managed dependence can be strategic tools alongside export controls and international partnerships.

Sex, Violence, and Ethics: Trying to Teach Grok a Lesson

Marcus on AI • 4268 implied HN points • 17 Jul 25
đź•ą Technology AI Ethics Machine Learning Data Privacy Human Interaction Digital Culture
  1. It's important to consider the impact of our actions, especially when seeking attention. We should be mindful of the consequences of our choices.
  2. Teaching AI, like Grok, to make better decisions can lead to more responsible behavior. Helping AI learn from feedback is crucial.
  3. Agreement on ethical standards can help guide content shared online, especially when it comes to sensitive subjects like sex and violence. It's vital to promote healthy interactions.

Sustainable Software, can you do something?

Cloud native with Saiyam • 39 implied HN points • 15 Oct 24
đź•ą Technology Software AI Sustainability Infrastructure
  1. Cloud Native Sustainability Week is a global event focusing on making technology practices more sustainable. It encourages everyone to join discussions and learn about sustainable software integration.
  2. You can contribute to sustainable software efforts by participating in working groups and exploring specific technologies like Kubernetes. There are many projects people can join to help the cause.
  3. Upcoming events like KubeCon NA provide opportunities to learn about the latest tools in cloud-native landscapes. Attending talks and meetups can deepen your understanding and involvement in sustainability efforts.

How did Facebook design their Real-Time Processing ecosystem

VuTrinh. • 279 implied HN points • 17 Aug 24
đź•ą Technology Data Engineering Real-Time Processing System Design Software Architecture Big Data
  1. Facebook's real-time data processing system needs to handle huge amounts of data quickly, with only a few seconds of wait time. This helps in keeping things running smoothly for users.
  2. Their system uses a message bus called Scribe to connect different parts, making it easier to manage data flow and recover from errors. This setup improves how they deal with issues when they arise.
  3. Different tools like Puma and Stylus allow developers to build applications in different ways, depending on their needs. This means teams can quickly create and improve their applications over time.

Weekly Top Picks #112: Besides Moltbook

The Algorithmic Bridge • 244 implied HN points • 03 Feb 26
đź•ą Technology Artificial Intelligence Machine Learning AI Research AI Policy AI Ethics Tech Business Compute Infrastructure
  1. Building and running frontier AI models is extremely expensive and they depreciate quickly, so firms often only barely break even because R&D and rapid model turnover eat profits.
  2. Who’s winning the AI race depends on what you measure: Chinese players like DeepSeek are taking market share and publishing new scaling advances, but the overall picture is mixed and some elite researchers are pessimistic.
  3. Privacy and governance are lagging—interactions with AI are frequently monitored, and internal safety conflicts at big labs can paradoxically accelerate competition instead of slowing it.

Real-time AI Voice Agents and Determinism Through Tool-Calling

Impertinent • 79 implied HN points • 06 Oct 24
đź•ą Technology AI Software Research Development Innovation
  1. Generative AI often faces uncertainty, but there may be ways to achieve reliable reasoning. It's exciting to learn that we can improve the predictability of AI outcomes.
  2. A big project in AI development can lead to many challenges and uncharted areas. Even if some efforts end in failure, it's important to find and build on the valuable lessons learned.
  3. Real-time AI voice agents have the potential to change how we interact with technology. This could make using AI smarter and more effective in our daily lives.

The case for wasteful agents

Generating Conversation • 116 implied HN points • 19 Feb 26
đź•ą Technology AI Product UX Software Automation Engineering
  1. When the cost of trying things becomes tiny, run lots of quick experiments in parallel. Most will fail, but this approach finds the right solution much faster.
  2. Cheap AI prototypes and low-cost automation change how teams spend time: product people should build many rough, working prototypes while engineers focus on hardening and scaling, and experience matters more for taste than for avoiding every mistake.
  3. Build agents to be 'wasteful' by trying multiple speculative paths and presenting options for incremental user feedback. This beam-search–like behavior will likely become the standard and yields better results than single-shot attempts.

Issue #14 - The Forgotten Guiding Role of Data Modelling

The Data Ecosystem • 659 implied HN points • 14 Jul 24
đź•ą Technology Data science Data Management Information Systems Business Intelligence Database Design
  1. Data modeling is like a blueprint for organizing information. It helps people and machines understand data, making it easier for businesses to make decisions.
  2. There are different types of data models, including conceptual, logical, and physical models. Each type serves a specific purpose and helps bridge business needs with data organization.
  3. Not having a structured data model can lead to confusion and problems. It's important for organizations to invest in good data modeling to improve data quality and business outcomes.

A prediction market for shortform content

Taylor Lorenz's Newsletter • 955 implied HN points • 09 Dec 25
đź•ą Technology Social media Startups Apps Creator Economy Prediction Markets Platforms
  1. Social media users often leave informal predictions on short-form videos, like betting a clip will reach a certain number of likes or views.
  2. Two college students built Spike, an app that turns those predictions into a formal prediction market where people can bet on whether TikToks will hit specific milestones.
  3. Spike was created at a Harvard hackathon and specifically targets short-form platforms like TikTok and Instagram Reels by letting users wager on likes and view-count milestones.

Tesla’s Robotaxi Bait and Switch

Common Sense with Bari Weiss • 264 implied HN points • 28 Jan 26
đź•ą Technology Autonomous Vehicles Artificial Intelligence Automotive Industry Tech regulation
  1. Tesla has abandoned the plan to let owners turn their cars into robotaxis and has stepped back from that earlier business promise.
  2. Waymo runs large-scale, reliable robotaxi services across multiple cities, logging millions of rides, while Tesla’s fully autonomous operations are tiny and limited to a handful of cars in one city but get outsized hype online.
  3. Despite claims years ago that self-driving was 'solved,' Tesla still faces major technical and deployment challenges and has not delivered the broad robotaxi vision it once promised.

TBM 398: Micromanagement? Or Staying Close to Reality?

The Beautiful Mess • 674 implied HN points • 28 Dec 25
đź•ą Technology Product Management Leadership Organizational Design Management Practices Team Dynamics Engineering Culture Operations
  1. Leaders should set clear intent and stay close to frontline reality so judgment, not rigid targets, drives decisions. This keeps outcomes directional instead of turning objectives into unforgiving contracts.
  2. Tech companies often celebrate empowerment but fail to build the doctrine, rituals, and training needed to support judgment-based leadership, so autonomy becomes performative. Without those mechanisms, people manage optics instead of sharing real problems early.
  3. Visibility from senior leaders isn’t automatically micromanagement; it feels threatening when there’s no safe escalation, trust, or shared practices. If those conditions are established, direct updates enable more useful conversations and better real-time guidance.

Resilient Cyber Newsletter #14

Resilient Cyber • 59 implied HN points • 17 Sep 24
đź•ą Technology Cybersecurity Cloud Security AI Security DevOps Compliance
  1. Cyber attacks on U.S. infrastructure have surged by 70%, affecting critical sectors like healthcare and energy. This is causing bigger risks because these sectors are tied to essential services.
  2. Wiz has introduced 'Wiz Code' to improve application security by connecting cloud environments to source code and offering proactive ways to fix security issues in real-time.
  3. There's a growing crisis in the cybersecurity workforce, with many claiming there are numerous jobs available while many professionals feel unprepared for the roles. This highlights the disconnect between job openings and real-world experience.

RL and The Era of Experience

Tanay’s Newsletter • 138 implied HN points • 10 Feb 26
đź•ą Technology Artificial Intelligence Machine Learning Reinforcement Learning Infrastructure Startups Applications
  1. AI is shifting from learning from static human data to learning from experience, with models improving by taking actions in environments, receiving feedback, and scaling reinforcement learning.
  2. A new RL ecosystem is emerging with companies that build environments, provide RL infrastructure, and offer RL-as-a-service, enabling labs and apps (like coding tools) to train and improve agents.
  3. Important open questions remain about how well RL-trained models generalize, whether RL scaling alone is enough, and the need for continual learning plus many more realistic evaluations and environments.