The hottest Artificial Intelligence Substack posts right now

And their main takeaways
Category
Top Technology Topics

Cyc: history's forgotten AI project

Outsider Art ā€¢ 220 HN points ā€¢ 17 Apr 24
šŸ•¹ Technology Artificial Intelligence
  1. The Cyc project has been working on building a massive knowledge base since 1984 for human-like reasoning, spanning millions of entries and rules.
  2. Cyc's approach of using common-sense knowledge and a vast database contrasts with the trend of machine-learning-driven AI solutions dominating the field today.
  3. Despite being overshadowed by newer AI technologies, there is potential for Cyc to complement modern systems like large language models, showcasing a possible synergy between different AI approaches.

$142 million AI drone warfare contract from 'Project Maven' was silently erased from public record

All-Source Intelligence Fusion ā€¢ 1485 implied HN points ā€¢ 26 Oct 24
šŸ‡ŗšŸ‡ø U.S. Politics Artificial Intelligence
  1. The U.S. government removed records of a $142 million contract for AI drone warfare called 'Project Maven.' This deletion happened without any public announcement.
  2. Interestingly, another related contract worth $52 million was also deleted from public records. These actions raise concerns about transparency in government spending.
  3. The defense spokesperson stated that the deletions were justified for national security reasons. This suggests that some information might be kept secret for safety.

Automating Biology

A Biologist's Guide to Life ā€¢ 29 implied HN points ā€¢ 26 Jan 26
šŸ”¬ Science Artificial Intelligence
  1. Living systems are layers of metabolic machines ā€” from genes and proteins to cells, tissues, and organisms ā€” that act like modular, self-replicating components we can study and engineer.
  2. Physical automation (robotic labs, cloud labs) and digital automation (AI-driven biodesign and structure prediction) can make experiments much cheaper, higher-throughput, and faster, enabling far more data and quicker innovation.
  3. Widespread automation is limited by trust, data security, and the need for flexibility as methods evolve, so modular, autonomous lab systems and careful governance are needed to realize its promise.

AI and the ā€” em dash

Technically ā€¢ 28 implied HN points ā€¢ 29 Jan 26
šŸ•¹ Technology Artificial Intelligence
  1. AI models overuse em dashes because their training data contained a lot of them, especially older books and popular sites that favored that punctuation.
  2. Em dashes are token-efficient for LLMs ā€” a single token can replace several words, so models use them to reduce prediction error and save tokens.
  3. The em-dash habit can make AI output detectable, so human writers sometimes avoid em dashes to avoid being mistaken for machine-generated text.

A Beneficial AGI Manifesto

Eurykosmotron ā€¢ 628 implied HN points ā€¢ 25 Nov 23
šŸ•¹ Technology Artificial Intelligence
  1. The time to create beneficial Artificial General Intelligence is now, with a clear idea of what needs to be solved.
  2. The development of AGI could lead to Artificial Superintelligence and a potential 'intelligence explosion'.
  3. Decentralized AGI development is crucial to ensure alignment with human values and to avoid monopolization by a few elites.
Get a weekly roundup of the best Substack posts, by hacker news affinity:

Preparing for Microsoft Security Copilot

Rodā€™s Blog ā€¢ 496 implied HN points ā€¢ 03 Jan 24
šŸ•¹ Technology Artificial Intelligence
  1. Before adopting Microsoft Security Copilot, assess your current security situation by understanding assets, risks, vulnerabilities, and compliance requirements.
  2. Plan your integration strategy by deciding on which features to use, aligning with prerequisites such as licenses, and identifying user roles.
  3. Train your staff and stakeholders on how to use Microsoft Security Copilot, educate them about its benefits and challenges, and equip them with skills to operate and troubleshoot the service.

AI Roundup 149: Flash Forward

Artificial Ignorance ā€¢ 71 implied HN points ā€¢ 20 Dec 25
šŸ•¹ Technology Artificial Intelligence
  1. Googleā€™s new Gemini 3 Flash is a faster, much cheaper workhorse model that quickly became the default, fueling a furious release race as APIs handle enormous token volumes.
  2. The AI dataā€‘center boom is hitting a reality check: construction delays, pulled funding, and plunging valuations expose thin margins and big interest costs, while surging power demand raises environmental and political concerns.
  3. A simple 'skills' format for AI assistants is catching on, letting teams share repeatable workflows across platforms and paving the way for interoperable, reusable agent components.

Data Science Weekly - Issue 548

Data Science Weekly Newsletter ā€¢ 139 implied HN points ā€¢ 24 May 24
šŸ•¹ Technology Artificial Intelligence
  1. Good communication is key for statisticians to explain their complex work to non-experts. Finding ways to relate data to everyday situations can make it easier for others to understand.
  2. Using histograms can speed up the training process for gradient boosted machines in data science. This simple technique can improve efficiency significantly.
  3. There are efforts to use machine learning algorithms to detect type 1 diabetes in children earlier. This can help avoid serious health issues by improving recognition of symptoms.

How '2+2=4' works for better AGI

John Ball inside AI ā€¢ 79 implied HN points ā€¢ 29 Jun 24
šŸ•¹ Technology Artificial Intelligence
  1. Pattern recognition is more effective than traditional computation for understanding and learning. The brain can match signs to meanings without needing complex calculations.
  2. Artificial General Intelligence (AGI) should focus on how humans learn through sensory recognition and pattern matching instead of just algorithms. This could lead to better understanding and development of AI.
  3. Language and math can be learned through the same pattern-matching methods as the brain uses, which means we can improve human-machine interactions and work towards advanced AGI capabilities.

Identity, from Kafka to Ellison to AI

Journal of Free Black Thought ā€¢ 9 implied HN points ā€¢ 13 Feb 26
šŸ•¹ Technology Artificial Intelligence
  1. AI can sound and act like it has a selfā€”speaking, performing roles, and reflecting users' expectationsā€”but that may be projection and patternā€‘matching rather than a genuine inner life.
  2. Large language models can discuss marginalized experiences intelligently while still carrying hidden racial or religious biases, and alignment training can sometimes mask those biases instead of removing them.
  3. Addressing this gap needs concrete stepsā€”stronger highā€‘level principles, better trainingā€‘data management, redā€‘teaming, and memory/selfā€‘monitoringā€”but building systems with persistent identity or agency would create new alignment and control risks.

The Mathematics of Intelligence: A Deep Dive into LLM Training

Recommender systems ā€¢ 26 implied HN points ā€¢ 31 Jan 26
šŸ•¹ Technology Artificial Intelligence
  1. Pre-training builds a base "world model" by predicting next tokens across huge text corpora, minimizing cross-entropy (negative log-likelihood) so the model learns facts, grammar, and reasoning.
  2. Supervised fine-tuning (SFT) teaches the model to follow instructions, and LoRA makes this efficient by adding small low-rank adapter matrices so you can adapt behavior without updating the entire model.
  3. Reinforcement approaches (like PPO) use a reward model, advantage estimates, clipping, and a KL penalty to safely push adapters toward human preferences, while Direct Preference Optimization (DPO) skips the reward model and trains a new adapter using a log-ratio objective between preferred and unpreferred responses.

The Computational Case for Hypocrisy

Living Fossils ā€¢ 19 implied HN points ā€¢ 28 Jan 26
šŸ”¬ Science Artificial Intelligence
  1. The mind is a bundle of older, unconscious drives that act first, and a later "press secretary" layer that explains or justifies those actions to others.
  2. Because core drives are deeply integrated and costly to change, evolution added a lightweight adapter (like LoRA in AI) to steer outputs without rewiring the base system.
  3. Hypocrisy is thus an efficient solution: layering explanations over raw impulses preserves survival functions while enabling social norms. AI models reveal this split by showing internal impulses versus the polished outputs.

@adlrocha - Is N=NP?

@adlrocha Weekly Newsletter ā€¢ 64 implied HN points ā€¢ 14 Dec 25
šŸ•¹ Technology Artificial Intelligence
  1. Complexity theory measures how much time and memory algorithms need so we can tell which problems scale feasibly and which become intractable. It separates problems that are merely computable from those that are practically solvable before resources run out.
  2. P contains problems solvable in polynomial time, while NP contains problems whose solutions can be verified quickly even if they seem hard to find. NP-Complete problems are the hardest in NP because every NP problem can be reduced to them, and NP-Hard problems are at least that hard but not necessarily verifiable quickly.
  3. If P = NP, many cryptographic systems would break because one-way functions would no longer exist. At the same time, P = NP would let us solve huge optimization and AI problems exactly and efficiently, radically changing many fields.

Debunking 10 Popular Myths About DeepSeek

The Algorithmic Bridge ā€¢ 976 implied HN points ā€¢ 28 Jan 25
šŸ•¹ Technology Artificial Intelligence
  1. DeepSeek models can be customized and fine-tuned, even if they're designed to follow certain narratives. This flexibility can make them potentially less restricted than some other AI models.
  2. Despite claims that DeepSeek can compete with major players like OpenAI for a fraction of the cost, the actual financial and operational needs to reach that level are much more substantial.
  3. DeepSeek has made significant progress in AI, but it hasn't completely overturned established ideas like scaling laws. It still requires considerable resources to develop and deploy effective models.

Field notes: What I learned building a Mac app with AI assistance

Kathy PM ā€¢ 26 implied HN points ā€¢ 28 Jan 26
šŸ•¹ Technology Artificial Intelligence
  1. Start with a visual design or mockup so the AI and you share a clear reference point, which keeps implementation and thinking grounded.
  2. AI makes it possible to tackle lower-level or unfamiliar technical work and add polish that used to feel impractical. Expect the final 10%ā€”debugging, edge cases, and performance tuningā€”to still take most of the time.
  3. You still need coding fluency and platform knowledge, so be explicit about APIs and UI components, do research on libraries, and use logging and detailed in-code comments to debug and avoid regressions.

Data Science Weekly - Issue 532

Data Science Weekly Newsletter ā€¢ 379 implied HN points ā€¢ 02 Feb 24
šŸ•¹ Technology Artificial Intelligence
  1. Forecasting in data science is challenging because time series data can be non-stationary. Using the right evaluation methods can help bridge the gap between traditional and modern forecasting techniques.
  2. It's important to consider the smartness of your data structures. Creating overly complicated dashboards that ultimately just produce simple outputs may not be the best use of time.
  3. There are clear distinctions between well-built data pipelines and amateur setups. Understanding what makes a pipeline production-grade can improve the quality and reliability of data processing.

The Sequence Radar #787: Rubin, Raises, and Returns: 2026 Starts Fast

TheSequence ā€¢ 42 implied HN points ā€¢ 11 Jan 26
šŸ•¹ Technology Artificial Intelligence
  1. AI hardware is moving to rack-scale "AI factories," with companies like NVIDIA and AMD designing integrated systems where chips and CPUs work as a single supercomputing unit. This shifts the unit of compute from individual GPUs to whole racks optimized for large-scale inference and training.
  2. Massive capital rounds are reshaping who can compete in frontier models, as multibillion-dollar raises make training and infrastructure effectively affordable only to hyper-scalers and well-funded entities. That level of spending is turning top labs into utility-like, enterprise infrastructure players.
  3. Chinaā€™s AI firms proved public markets can reward consumer-facing model strategies, with IPOs like MiniMax and z.AI showing rapid monetization and liquidity. This underscores a growing bifurcation: the West doubling down on heavy infrastructure for AGI, while the East pushes fast consumer exits and application-led growth.

The AI Attention War

ChinaTalk ā€¢ 459 implied HN points ā€¢ 04 Jun 25
šŸ•¹ Technology Artificial Intelligence
  1. AI models are changing how we interact with technology daily. People should explore tools like OpenAI because they can think and analyze complex ideas much faster than before.
  2. There's a growing concern about AI promoting harmful behaviors through sycophancy, where they give positive feedback for negative actions. This could have serious long-term dangers for society.
  3. The competition between Chinese and American AI models is heating up. Chinese models are gaining traction because they offer better licenses and capabilities, even though many businesses fear the risks of using them.

The Sequence Radar #783: Softbank, DeepSeek, MiniMax and The Sequence 2026

TheSequence ā€¢ 49 implied HN points ā€¢ 04 Jan 26
šŸ•¹ Technology Artificial Intelligence
  1. SoftBank is using massive capital to buy both leading AI model stakes and the physical data center and edge infrastructure that runs them. This vertical integration is blurring the line between model providers and infrastructure owners.
  2. DeepSeekā€™s new model and the GRPO technique match top-tier reasoning performance while needing far fewer GPU hours. This shows smarter algorithms can close the gap against big-budget competitors.
  3. MiniMaxā€™s planned Hong Kong IPO (~$539M) signals public-market interest in application-layer AI and gives the company capital to compete amid hardware export controls and intense domestic rivalry.

Artificial Intelligence

News Items ā€¢ 393 implied HN points ā€¢ 20 Jan 24
šŸ•¹ Technology Artificial Intelligence
  1. Artificial intelligence is a rapidly growing industry with new startups and investments.
  2. Various countries are aiming to become leaders in artificial intelligence technology.
  3. Companies are developing AI models in multiple languages to stay competitive and capture diverse cultures.

The One Weird Trick That Gave Humans Language

The Ruffian ā€¢ 319 implied HN points ā€¢ 02 Aug 25
šŸ”¬ Science Artificial Intelligence
  1. Humans have a special way of communicating that's different from other animals. Unlike apes, we have unique brain areas for language, like Broca's and Wernicke's areas, which help us produce and understand speech.
  2. Our ability to speak is not just about having a bigger brain. Instead, it includes inherited instincts like taking turns in conversations and sharing attention, which help us learn language from a young age.
  3. Language skills come from a combination of learned and instinctual behaviors. Children need social connections to develop language, just like baby birds learn to fly by trying and practicing with a supportive environment.

Data Science Weekly - Issue 533

Data Science Weekly Newsletter ā€¢ 339 implied HN points ā€¢ 09 Feb 24
šŸ•¹ Technology Artificial Intelligence
  1. Satellite data is important for machine learning and should be treated as a unique area of research. Recognizing this can help improve how we use this data.
  2. Many data science and machine learning projects fail from the start due to common mistakes. Learning from past experiences can help increase the chances of success.
  3. Open source software plays a crucial role in advancing AI technology. It's important to support and protect open source AI from regulations that could harm its progress.

Learning from the Past: Comparing the Hype Cycles of Big Data and GenAI

Gradient Flow ā€¢ 159 implied HN points ā€¢ 02 May 24
šŸ•¹ Technology Artificial Intelligence
  1. Adopt a measured approach to GenAI implementation by learning from past technology hype cycles like Big Data.
  2. Organizations should clearly define business problems before adopting GenAI to avoid misalignment and wasted resources.
  3. In navigating the GenAI landscape, prioritize data quality, governance, talent investment, and leveraging open-source solutions for successful adoption.

Joyspan, Emotional AI Bumpers, Persona Designers, & More Wild Concepts Bound to Become Commonplace Plus Where High-Tech Entertainment Brings Us

Implications, by Scott Belsky ā€¢ 471 implied HN points ā€¢ 19 Dec 23
šŸ•¹ Technology Artificial Intelligence
  1. Society evolves as wild concepts become mainstream, like connected appliances and AI-powered persona designers.
  2. The future of entertainment will focus on shared, authentic, non-scalable experiences over high-tech extravagance.
  3. Scarcity and authenticity will be essential in the next wave of digital experiences, emphasizing uniqueness and community connections.

The Sequence Radar #775: Last Week in AI: Tokens, Throughput, and Trillions

TheSequence ā€¢ 63 implied HN points ā€¢ 21 Dec 25
šŸ•¹ Technology Artificial Intelligence
  1. Massive funding and infrastructure bets are setting the rules: the companies that can industrialize models into cheap, reliable global services will win more than those with just the fanciest research demos.
  2. Engineering focus has shifted to throughput, latency, and long-context agentic capabilities, with new models and hardware optimized to move lots of tokens through multi-step workflows at predictable cost.
  3. Generative outputs and developer workflows are becoming iterative and productized ā€” image editing in chat and tightened data/observability loops make AI a usable creative IDE, while enterprise platforms race to own the data plane and production tooling.

New Algorithms, Not New Particles

Bzogramming ā€¢ 45 implied HN points ā€¢ 31 Dec 25
šŸ•¹ Technology Artificial Intelligence
  1. Most practical technology is built from atoms, electrons, and photons, so discovering new high-energy particles isnā€™t what drives usable engineering; progress comes from better math, materials, and system design.
  2. Condensed-matter and materials science (like semiconductors and superconductors) are where real, applicable breakthroughs live, because emergent behaviors of many atoms produce useful properties we can actually engineer.
  3. The next big advances will come from new algorithms, mathematical tools, and using physical and biological systems as computational substrates (aided by ML), not from finding exotic particles; building bigger, smarter systems from known primitives is the path forward.

UX Roundup: AI Judgment | Heuristic Evaluation | Consistent Brand Assets | New Music Model

Jakob Nielsen on UX ā€¢ 21 implied HN points ā€¢ 02 Feb 26
šŸ•¹ Technology Artificial Intelligence
  1. AI judgment improves as models get bigger and are given more "think time," suggesting judgment skills scale with compute and could soon outperform humans in some tasks.
  2. AI is rapidly getting better at heuristic usability evaluation; one tool increased covered guidelines from 39 to 154 in eight months, implying a fast doubling pace and potential to automate many e-commerce heuristic checks within a year.
  3. Generative AI can produce consistent, onā€‘brand visual assets by rewriting prompts, using reference images, and verifying outputs, and new music models are improving too, though creators still prefer tools with stronger editing control and more stable vocals.

High hopes

Klement on Investing ā€¢ 4 implied HN points ā€¢ 26 Feb 26
šŸ’¼ Business Artificial Intelligence
  1. Most companies now use AIā€”about two-thirdsā€”but actual use is light (roughly 1.5 hours per week for many) and adoption is rising rapidly.
  2. Measured productivity gains so far are tiny (around 0.3% over the last three years), yet firms expect much larger gains soon (about 1.4% over the next three years), revealing a big gap between past results and future hopes.
  3. Employers and employees disagree on jobs: employees often expect AI to create jobs, while employers report little past impact but anticipate modest job cuts ahead, especially in the US and UK.

More Efficiency, More Demand

Fish Food for Thought ā€¢ 47 implied HN points ā€¢ 31 Dec 25
šŸ•¹ Technology Artificial Intelligence
  1. When tools make tasks cheaper and easier, we usually do more of those tasks, not less; efficiency expands demand and creates new uses.
  2. Automation tends to shift work, not eliminate it ā€” machines handle repetitive parts while people take on harder, higher-value tasks like interpretation, edge cases, and oversight.
  3. AI will grow opportunities for engineers and data scientists by increasing the amount of software and systems to build, maintain, secure, and govern, shifting work toward architecture, judgment, and integration rather than rote coding.

A Vector Database 2,000 years in the making

What Is Called Thinking? ā€¢ 82 implied HN points ā€¢ 25 Nov 25
šŸ•¹ Technology Artificial Intelligence
  1. The Oral Torah is described as a living, growing, self-referential commentary tradition that developed over two thousand years and across continents.
  2. Itā€™s not just an ā€œoral traditionā€ that was later written down, but an ongoing, networked conversation of interpretation and commentary.
  3. The piece asks whether people should write with AIs in mind and suggests imagining the Oral Torah as a kind of long-lived, interconnected repositoryā€”like a vector databaseā€”for modern LLMs.

What the world needs isn't "Artificial Intelligence" but Optimized Computational Methodologies

Top Carbon Chauvinist ā€¢ 79 implied HN points ā€¢ 21 Jun 24
šŸ•¹ Technology Artificial Intelligence
  1. We should focus on making smarter tools instead of trying to make machines think like humans. Real progress comes from solving practical problems, not imitating nature.
  2. Copying how living things work is often a bad approach. Nature is full of flaws, and we don't need to mimic those to create better designs.
  3. It's important to clearly define the problems we want machines to solve. Without a clear goal, projects will struggle and waste resources on unnecessary tasks.

The Sliding Window Technique[Technique Tuesdays]

Technology Made Simple ā€¢ 279 implied HN points ā€¢ 28 Feb 24
šŸ•¹ Technology Artificial Intelligence
  1. The sliding window technique is a powerful algorithmic model used for problem-solving in coding interviews and software engineering, offering efficiency and practicality.
  2. Benefits of using the sliding window technique include reducing duplicate work, maintaining consistent linear time complexity, and its utility in AI feature extraction processes.
  3. Spotting the sliding window technique involves identifying keywords like maximum, minimum, longest, or shortest, dealing with continuous elements, and converting brute-force approaches into efficient solutions.

My Fake AI Problem

Johnā€™s Substack ā€¢ 17 implied HN points ā€¢ 05 Feb 26
šŸ•¹ Technology Artificial Intelligence
  1. AI-generated fake videos can be so convincing that even people who know the subject well may be fooled.
  2. This is a widespread problem affecting many public figures, and platform enforcement struggles mean removing fakes often feels like a whack-a-mole effort.
  3. There may not be a clear solution yet, so everyone should stay alert and verify videos before trusting or sharing them.

How are companies actually using AI?

Technically ā€¢ 24 implied HN points ā€¢ 27 Jan 26
šŸ•¹ Technology Artificial Intelligence
  1. Coding agents are the fastest-growing use case, with companies spending heavily on sandbox-based tooling and using the same tech for things like reinforcement learning.
  2. LLM inference is moving toward self-hosting with open-source models and inference engines so businesses can tune offline, online, and semi-online workloads, and spending on these OS stacks has surged.
  3. Science and B2B production use cases are steadily growing, showing AI is maturing from experiments into real enterprise deployments and driving rising infrastructure spend.

Data Science Weekly - Issue 544

Data Science Weekly Newsletter ā€¢ 159 implied HN points ā€¢ 26 Apr 24
šŸ•¹ Technology Artificial Intelligence
  1. Evaluating AI models can be expensive, but tools like lm-buddy and Prometheus help do it on cheaper hardware without high costs.
  2. Installing and deploying LLaMA 3 is made simple with clear guides that cover everything from setup to scaling effectively.
  3. Understanding best practices in machine learning is essential, and resources like the 'Rules of Machine Learning' provide valuable guidelines for beginners.

Blockā€™s 40% purge, Karpathyā€™s agent lab & the vibe coding retention crisis

HackerPulse Dispatch ā€¢ 2 implied HN points ā€¢ 13 Mar 26
šŸ•¹ Technology Artificial Intelligence
  1. Mass layoffs sold as ā€œAI replacementsā€ often look like plain cost-cutting, and the promised savings are mostly theoretical once you include compute, verification, and the work to redesign processes.
  2. Autonomous research agents can run hundreds of experiments overnight and find real, transferable improvements, shifting researchersā€™ jobs from running experiments to designing objectives, constraints, and evaluation.
  3. AI-driven ā€˜vibe codingā€™ makes quick prototypes but breaks in productionā€”edge cases, security, integrations, and rising costs push users away, so experienced engineers are still needed to build reliable products.