The hottest Artificial Intelligence Substack posts right now

And their main takeaways
Category
Top Technology Topics

2025 Open Models Year in Review

Democratizing Automation • 292 implied HN points • 14 Dec 25
đź•ą Technology Artificial Intelligence
  1. Open models made a dramatic jump in 2025, matching closed models on many benchmarks and becoming realistic options for real-world deployments beyond just privacy or fine-tuning.
  2. A few breakout releases — notably DeepSeek R1, Qwen 3, and Kimi K2 — had outsized influence, driving wider adoption and encouraging more open licensing from major labs, especially in China.
  3. The ecosystem exploded in scale and variety, with thousands of new models uploaded monthly, clear specialist niches and a public tiering of makers, leaving open models established and poised for further growth in 2026.

Augmented Coding: Beyond the Vibes

Software Design: Tidy First? • 1855 implied HN points • 25 Jun 25
đź•ą Technology Artificial Intelligence
  1. Augmented coding is different from vibe coding. It's about caring for the code quality and complexity, not just getting the system to work.
  2. Keeping the project scope clear is key. You should focus on specific tasks, like creating a B+ Tree, while ensuring the code is tidy and functional.
  3. Collaboration with AI tools can enhance coding efficiency. You can rely on AI for tasks like writing tests or suggesting optimizations, but you must guide it to stay on track.

MIT Study: Using ChatGPT Won't Make You Dumb (Unless You Do It Wrong)

The Algorithmic Bridge • 1942 implied HN points • 19 Jun 25
đź•ą Technology Artificial Intelligence
  1. Using AI tools like ChatGPT can make you less engaged mentally if used excessively. People can become reliant on these tools and stop thinking deeply.
  2. When people switch from using AI tools back to using their own knowledge, they can struggle at first but may learn and grow better in the long run.
  3. The best way to use AI is to first work on a task with your own skills and then use AI to enhance what you've done, rather than relying on it from the start.

AI influencer hype gone wild

Marcus on AI • 5572 implied HN points • 31 Oct 24
đź•ą Technology Artificial Intelligence
  1. Many people are trying AI tools, but not everyone thinks they are effective. This shows there's a mix of interest and skepticism in using new technology.
  2. A recent survey revealed that while 79% of people have tried Microsoft Copilot, only 25% found it worthwhile. This indicates people are testing AI but still unsure about its overall value.
  3. People are not ignoring AI; they are being cautious and waiting to see if it meets their expectations before fully committing. It’s a wait-and-see attitude towards technology.

If You've Been Thinking About Subscribing to Premium...

Software Design: Tidy First? • 397 implied HN points • 22 Nov 25
đź•ą Technology Artificial Intelligence
  1. Limited-time Black Friday deal: $180/year through December 1st, reduced from the usual $250.
  2. Paid subscribers get early access to unpolished essays, a problem-solving chat community, and weekly "Thinkies" that teach habits for creative thinking.
  3. The project aims to help technical people feel safer as machines start to code, exploring responsibility and what changes when capabilities and speed increase.
Get a weekly roundup of the best Substack posts, by hacker news affinity:

How I code going into 2026

Olshansky's Newsletter • 183 implied HN points • 05 Jan 26
đź•ą Technology Artificial Intelligence
  1. Most coding is now delegated to AI agents, so engineers spend their time orchestrating agent personalities and guiding work rather than writing code by hand.
  2. Practical workflows matter: use Makefiles as a stable CLI, leave TODOs instead of side quests, maintain prompts/skills, write short copy-paste friendly docs, and review critical diffs on GitHub.
  3. Team roles and skills are shifting: leaders must be hands-on translators of intent into agent-driven work, focusing on system design, taste, and continuously improving agent behavior.

poe's law comes into full flower

bad cattitude • 223 implied HN points • 18 Dec 25
đź•ą Technology Artificial Intelligence
  1. AI can now create fake people and media so convincing that ordinary people can’t tell what’s real, blurring the line between parody and reality.
  2. That breakdown of trust will upend industries and enable widespread fraud and misinformation, while existing detection and verification tools are losing the arms race.
  3. A possible upside is that people and businesses may return to high-trust, in-person local interactions and city centers, which could revive communities and improve wellbeing.

We Interrupt the Ongoing Crisis for a Smidgen of Good News

The Crucial Years • 3677 implied HN points • 29 Jan 25
🌞 Climate & Environment Artificial Intelligence
  1. The new Chinese AI program DeepSeek uses only a small fraction of the electricity needed by similar American AI systems. This could challenge the fossil fuel industry's excuse for building more power plants based on increased energy demands from AI.
  2. Fossil fuel stocks have not been performing well in comparison to the broader market for several years, raising concerns about the industry's future in a world moving towards decarbonization.
  3. In Europe, solar energy has recently outperformed coal for the first time, marking a significant shift towards renewable energy sources in the region.

The new AI scaling law shell game

Marcus on AI • 4663 implied HN points • 24 Nov 24
đź•ą Technology Artificial Intelligence
  1. Scaling laws in AI aren't as reliable as people once thought. They're more like general ideas that can change, rather than hard rules.
  2. The new approach to scaling, which focuses on how long you train a model, can be costly and doesn't always work better for all problems.
  3. Instead of just trying to make existing models bigger or longer-lasting, the field needs fresh ideas and innovations to improve AI.

Latest open artifacts (#17): NVIDIA, Arcee, Minimax, DeepSeek, Z.ai and others close an eventful year on a high note

Democratizing Automation • 150 implied HN points • 05 Jan 26
đź•ą Technology Artificial Intelligence
  1. Several major open models and updates landed at year-end — releases from NVIDIA, Arcee, LLM360, Zhipu and others noticeably pushed open-model capabilities higher.
  2. The community trend is toward bigger and Mixture-of-Experts (MoE) architectures, multi-token prediction, and openly releasing training data and checkpoints, which should speed progress and reproducibility.
  3. Important tradeoffs remain: some models excel on specific tasks like UI or coding but can be slower or weaker on very long-context workloads, and even larger, more capable variants are promised in 2026.

2026 AI Predictions

Alex Ghiculescu's Newsletter • 135 implied HN points • 19 Jan 26
đź•ą Technology Artificial Intelligence
  1. AI labs will focus on coding agents, with most development effort and revenue moving toward models that write software.
  2. Keeping up with rapidly improving AI coding tools will be the main challenge for software companies; engineering teams will need to learn new workflows and roll them out across people with different skills and enthusiasm.
  3. New techniques will close agents' domain-knowledge gaps so models can understand real codebases and make decisions, and those same solutions will boost many other AI applications.

20 Predictions for AI in 2026

The Algorithmic Bridge • 286 implied HN points • 12 Dec 25
đź•ą Technology Artificial Intelligence
  1. A clear set of twenty specific predictions about how AI will develop in 2026 is presented.
  2. The piece reviews results from 2025 predictions and commits to being more specific and accountable to improve forecasting accuracy.
  3. Full access to the detailed content is behind a subscription paywall, though a 7-day free trial is offered.

The Sequence Radar #795: The New Inference Kids

TheSequence • 112 implied HN points • 25 Jan 26
đź•ą Technology Artificial Intelligence
  1. Serving models (inference) is now the main battleground, drawing huge funding as startups race to make model serving boring, reliable, and infinitely scalable.
  2. New kernel-level tricks are cutting recomputation and memory waste: RadixAttention reuses KV cache blocks like an LRU to avoid recomputing prefixes, and PagedAttention pages KV memory so GPUs can pack many more requests without VRAM fragmentation.
  3. Latency and per-turn cost now define product quality, causing a split in the stack between orchestration/hardware layers that manage scale and kernel teams that squeeze every FLOP to make models fast and cheap.

AlphaGeometry2: Impressive accomplishment, but still a long path ahead

Marcus on AI • 3161 implied HN points • 17 Feb 25
đź•ą Technology Artificial Intelligence
  1. AlphaGeometry2 is a specialized AI designed specifically for solving tough geometry problems, unlike general chatbots that tackle various types of questions. This means it's really good at what it was built for, but not much else.
  2. The system's impressive 84% success rate comes with a catch: it only achieves this after converting problems into a special math format first. Without this initial help, the success rate drops significantly.
  3. While AlphaGeometry2 shows promising advancements in AI problem-solving, it still struggles with many basic geometry concepts, highlighting that there's a long way to go before it can match high school students' understanding in geometry.

UX Roundup: AI is the Computer | Disrupting Filmed Entertainment | UX Trends | GOMS | Nano Banana 2 | Lyria 3

Jakob Nielsen on UX • 21 implied HN points • 02 Mar 26
đź•ą Technology Artificial Intelligence
  1. AI is becoming the computer itself: many specialized models will be orchestrated into a single, personal system that works on users' behalf and reduces the role of traditional user interfaces. This orchestration combines file systems, secure code execution, web access, and persistent memory to deliver personalized, autonomous capabilities.
  2. AI will disrupt filmed entertainment by improving production workflows, enabling small creators to produce professional-grade content, and spawning entirely new formats and distribution channels. These shifts could redirect tens of billions in industry revenue and reshape how audiences and legacy studios operate.
  3. AI is changing UX practice and tooling: models now make formal methods like GOMS cheap and practical for optimizing skilled-user efficiency, while new models (e.g., Nano Banana 2, Lyria 3) show steady progress but still have limits. Image generation is improving incrementally and music models remain short and constrained by copyright safeguards, so the tools are powerful but not yet perfect.

How much agency do we actually want?

benn.substack • 1585 implied HN points • 13 Jun 25
đź•ą Technology Artificial Intelligence
  1. Many people want clear directions to reach their goals rather than complete freedom to decide everything on their own. It's sometimes easier to follow a checklist than to choose your own path.
  2. In the tech world, even highly skilled professionals often seek specific instructions on what to do next, rather than relying solely on their creativity and initiative.
  3. While we talk about wanting more agency and independence, many of us really just want someone to give us a roadmap for success, even if it means giving up some of our freedom.

Everyone is excited about Claude Code!

Alex's Personal Blog • 164 implied HN points • 06 Jan 26
đź•ą Technology Artificial Intelligence
  1. Claude Code is giving lots of people superpowers by making it easy for non-developers and developers to build and ship useful software, democratizing who can create with AI.
  2. Nvidia’s new Vera Rubin chip suite and yearly upgrade push aim to satisfy booming AI compute demand and keep customers upgrading, but that strategy could still lead to a future chip glut and tougher price competition.
  3. U.S. moves toward Venezuela and talk about Greenland risk straining alliances and reshaping global tech markets, which could open opportunities for European and other non-U.S. tech companies.

Trust Networks Are the Antidote to AI Hype

Marginally Compelling • 15 implied HN points • 26 Feb 26
đź•ą Technology Artificial Intelligence
  1. Local AI agents that run on your machine and can access files and services feel magical but are still immature and can cause serious security and control failures.
  2. The AI news wave is overloaded with sensational claims, influencers, and speculative pieces that often mislead people and can even move markets without solid evidence.
  3. The best defense is a network of trusted, experienced people who actually test tools and do the hard work. Rely on them to soberly explain limits and filter the hype.

Maybe we should make ChatGPT less “WEIRD”

Sex and the State • 23 implied HN points • 19 Feb 26
đź•ą Technology Artificial Intelligence
  1. Large language models learn mainly from online content produced by Western, educated, industrialized, rich, democratic (WEIRD) populations, so their outputs reflect those perspectives more than the global population.
  2. WEIRD modes of thinking — more individualistic, analytical, and universalist — differ from many non-WEIRD, more holistic and group-focused cultures, which makes models less accurate or relevant for those other groups.
  3. That WEIRD bias can shape real-world effects: by reinforcing individualistic and commercial norms, LLMs may worsen loneliness and reduce real-world socializing with heavy use and advertising, so we should consider making models less WEIRD and study these downstream impacts.

The startups are thriving

Alex's Personal Blog • 262 implied HN points • 12 Dec 25
đź’Ľ Business Artificial Intelligence
  1. The federal government moved to preempt state AI laws by creating a task force and directing agencies to build a uniform national AI policy that can challenge conflicting state rules.
  2. A coalition of allied countries is coordinating to secure AI supply chains—investing in chips, rare earths, and infrastructure to reduce reliance on strategic rivals.
  3. AI-first startups are growing far faster than traditional benchmarks, posting huge ARR gains and forcing investors to expect growth well beyond the old T2D3 model.

Data Science Weekly - Issue 530

Data Science Weekly Newsletter • 1418 implied HN points • 19 Jan 24
đź•ą Technology Artificial Intelligence
  1. Good data visualization is important. Some types of graphs can be misleading, and it's better to avoid them.
  2. In healthcare, it's not just about having advanced technology like AI. The real focus should be on getting effective results from these technologies.
  3. Netflix released a lot of data about what people watched in 2023. Analyzing this can help us understand trends in streaming better.

DeepSeek Is Chinese But Its AI Models Are From Another Planet

The Algorithmic Bridge • 3344 implied HN points • 21 Jan 25
đź•ą Technology Artificial Intelligence
  1. DeepSeek, a Chinese AI company, has quickly created competitive AI models that are open-source and cheap. This challenges the idea that the U.S. has a clear lead in AI technology.
  2. Their new model, R1, is comparable to OpenAI's best models, showcasing that they can produce high-quality AI without the same resources. It suggests they might be using innovative methods to build these models efficiently.
  3. DeepSeek’s approach also includes letting their model learn on its own without much human guidance, raising questions about what future AI could look like and how it might think differently than humans.

Meta Buys Manus: Shifting Currents at Meta

Enterprise AI Trends • 168 implied HN points • 30 Dec 25
đź•ą Technology Artificial Intelligence
  1. Meta's acquisition of Manus rescues a fast-growing but unprofitable startup and rewards its founders and investors, while adding geopolitical and competitive implications.
  2. Because Manus relied heavily on Anthropic's Claude, the deal creates strategic tension — Meta could replace Claude in Manus's agent loop and become a direct competitor to Anthropic.
  3. The purchase highlights a bigger industry debate: Meta is betting that agent scaffolding and tools — not just foundational models — hold the most value, a stance that could reshape AI strategy and competition.

Something Big Is Happening in Cybersecurity

The Security Industry • 35 implied HN points • 17 Feb 26
đź•ą Technology Artificial Intelligence
  1. AI development is accelerating fast, with new models that feel like a qualitative leap and are even being used to build the next generation of models.
  2. The AI security market has exploded into hundreds of companies, including many focused on automating SOC work, and it has attracted substantial venture funding.
  3. AI security is becoming a standard part of organizational defenses, and soon it will no longer make sense to treat it as a separate category because every vendor will have AI-driven security features.

Top 10 Books we read in 2025

Graphs For Science • 105 implied HN points • 10 Jan 26
đź•ą Technology Artificial Intelligence
  1. A strong theme is practical engineering: many books show how to turn LLM demos into working agents using RAG, embeddings, knowledge graphs, tool use, and prompt patterns to make outputs more reliable and auditable.
  2. There’s a clear focus on hands-on playbooks and trade-offs—quick-starts, checklists, code examples, and patterns for prototyping, retrieval, latency/cost decisions, multi-agent orchestration, and production concerns.
  3. The collection balances technical how-to guidance with broader perspectives on responsible use, human uniqueness, organizational strategy, and interdisciplinary science, highlighting ethics, norms for academics, and big-picture questions about life and intelligence.

Abundance, Dislocation, & the Fight to Focus Our Attention in the 21st Century

Brad DeLong's Grasping Reality • 230 implied HN points • 10 Dec 25
đź•ą Technology Artificial Intelligence
  1. Material abundance has largely ended mass scarcity and improved health and longevity, but it doesn’t automatically give people meaning or a sense of agency; we must use wealth to create conditions for living wisely and well.
  2. Rapid technological change brings big gains but also disruptive dislocation and is being handled only moderately well by current politics. The emerging Info‑Bio‑Tech era makes attention the scarcest resource, so guarding focus against platform-driven capture is essential.
  3. The center of global growth is shifting toward the developing world, and the main political task is building institutions that expand real freedom—agency, dignity, and a shared sense of reality—so people can truly flourish.

2025 Interconnects year in review

Democratizing Automation • 195 implied HN points • 18 Dec 25
đź•ą Technology Artificial Intelligence
  1. The publication grew a lot this year and became a much more influential source of cutting‑edge AI analysis, reaching millions of pageviews and a much larger audience.
  2. Reinforcement learning, reasoning models, and open‑model ecosystems were the central technical themes, and major initiatives were launched to advance American open models and research infrastructure.
  3. Output hit practical limits after a year of high volume, so the focus is shifting to higher‑value work: prioritizing quality over quantity, investing in key projects, and using more open models going forward.

Nobody is shipping your agent’s code (yet) | Predictions from LinearB’s Ori Keren

Dev Interrupted • 56 implied HN points • 03 Feb 26
đź•ą Technology Artificial Intelligence
  1. AI has erased the blank-page problem and speeds up code generation, but those upstream gains are being lost to chaotic code reviews, testing, and integration unless teams build proper infrastructure.
  2. Agentic tools that can control your local machine (like OpenClaw/Moltbot) show huge power but create major security and governance risks, so most organizations won’t give them autonomous control yet.
  3. The economics of software are shifting: survival favors substrate-efficient tools and firms with unique data or "insight compression," and the current "dark flow" of vibe coding can make teams feel faster while actually introducing hidden bugs, so risk-aware pipelines and better testing are essential.

Recursive self-improvement

Metacritic Capital • 6 implied HN points • 10 Mar 26
đź•ą Technology Artificial Intelligence
  1. AI training and inference costs are falling rapidly, with practical community optimizations already cutting costs by large orders of magnitude.
  2. Cheaper models let you run far more reasoning tokens, and that extra compute predictably improves performance; reinforcement learning with verifiable rewards can crystallize those gains.
  3. Falling costs combined with inference-time scaling and agent swarms create a feedback loop that can drive recursive self-improvement, so investors should expect faster capability growth and significant economic and safety implications.

The Sequence Opinion #798: Inside the Most Important Paper in Agentic Reasoning: Why the Loop is the Logic

TheSequence • 84 implied HN points • 29 Jan 26
đź•ą Technology Artificial Intelligence
  1. Reasoning comes from the interaction loop with the environment, not just from the model itself.
  2. Current LLMs act like fast, shallow 'System 1' pattern matchers, so they need agentic feedback loops to produce real-world reasoning and agency.
  3. The next frontier is designing the agentic loop and environment (the "new hidden layer") rather than only scaling model parameters.

Casey (Newton) Strikes Out

Marcus on AI • 3517 implied HN points • 11 Dec 24
đź•ą Technology Artificial Intelligence
  1. AI skeptics believe that while there were big improvements in AI, those gains seem to be slowing down now. They think the hype isn't matching reality.
  2. Casey Newton's view oversimplifies AI skepticism by dividing it into two groups, but many skeptics have different opinions and concerns about AI's influence.
  3. It's important to recognize the problems with AI and financial issues in the industry, rather than just celebrating advancements without addressing weaknesses.

All That Is Solid Melts into Code

Daniel Pinchbeck’s Newsletter • 29 implied HN points • 14 Feb 26
đź•ą Technology Artificial Intelligence
  1. AI has reached an inflection point where models can rapidly automate broad white‑collar cognitive tasks. This is already eroding entry‑level jobs and changing roles like software engineers into architects and debuggers.
  2. If human labor becomes optional, the economy could see extreme wealth concentration and mass unemployment unless we redesign how abundance and income are shared. Without policy changes, the link between work and survival may break for many people.
  3. Powerful self‑improving AI brings huge opportunities—faster creativity and the collapse of old knowledge hierarchies—but also serious risks like cyberattacks or engineered harms, so urgent governance and planning are needed.

Pity the Harvard Undergrad

In My Tribe • 865 implied HN points • 07 Aug 25
đź’Ľ Business Artificial Intelligence
  1. AI is quickly taking over jobs that used to be done by humans, especially in fields like law and finance. This means fewer entry-level jobs for new graduates.
  2. Harvard graduates may need to find jobs that mix different skills, like working with people and technology, to stay relevant and employed.
  3. In the future, almost all jobs that rely on writing or analysis will involve software development. Graduates will need to think like software developers to stay valuable in the job market.

This Week in Marketing: Social Media Addiction vs. Human-Free Social Media and AI Layoffs

The Social Juice • 63 implied HN points • 01 Feb 26
đź’Ľ Business Artificial Intelligence
  1. Social platforms are in flux as users, creators and advertisers react to trust, moderation and product changes — some people are ditching apps like TikTok while new, AI‑only social networks and 'desocialized' feeds emerge.
  2. AI is reshaping media and jobs: companies are pouring money into agentic tools and ad tech even as some firms cut roles and many new AI startups and features debut, with uneven product success.
  3. Safety, legal and privacy pressures are rising as regulators, courts and publishers push back — youth addiction trials, encryption and data investigations, deepfakes and mass breaches are driving demands for controls and opt‑outs.

Will AI kill SaaS? The case for and against disruption

Brick by Brick • 45 implied HN points • 03 Feb 26
đź•ą Technology Artificial Intelligence
  1. AI that generates code and autonomous agents is collapsing the upfront cost of building software and can replace much of the human labor that SaaS products currently coordinate, threatening the old SaaS economic model.
  2. Big frictions—like high switching costs, regulatory and accountability needs, data gravity, and organizational inertia—make wholesale replacement of incumbent SaaS slow and hard.
  3. Disruption will be uneven and gradual: tools that automate repetitive, text-heavy workflows are most at risk, and winners will be challengers who target high-toil use cases or incumbents who proactively adopt agentic solutions.

The AI Manager's Schedule

Artificial Ignorance • 105 implied HN points • 16 Jan 26
đź•ą Technology Artificial Intelligence
  1. AI turns many maker tasks into delegated work, so your day shifts from long deep blocks to lots of short five-to-fifteen minute management intervals and juggling multiple agents.
  2. New top skills are clear vision, smart delegation, and orchestration — you need to know the end state, break work into bite-sized chunks, and run or coordinate multiple agents, and you must keep strong taste and bullshit detection to judge AI output.
  3. The change can speed up shipping and hugely amplify experienced people, but it also brings risks like micromanagement fatigue, juniors not learning, and initial slowdowns from debugging AI output; over time tools should reduce overhead and make these managerial skills broadly valuable.

LLMs - Part 1: Tokenization and Embeddings

Vasu’s Newsletter • 104 implied HN points • 05 Jan 26
đź•ą Technology Artificial Intelligence
  1. Text is split into discrete tokens, often subwords using Byte Pair Encoding, so a fixed vocabulary can represent any input by keeping common words whole and breaking rare words into parts.
  2. Each token ID is looked up in a learned embedding matrix to produce a dense vector, and these embeddings capture semantic and syntactic relationships learned during training.
  3. Embeddings are context-free and don’t encode position by themselves, so transformer mechanisms like attention and positional encodings combine them to determine meaning and word order.