Willingness to look stupid

Hacker News Comments

TLDR (Generated)

The “McDonald’s Theory” suggests that pitching a deliberately bad idea is the fastest way to overcome creative blocks and inspire a team to come up with better solutions.

• Break the ice with bad ideas: Suggesting something universally disliked (like McDonald’s for lunch) unites people in disagreement and forces them to generate superior alternatives.
• The first step is the hardest: Don’t overthink it; the primary goal is just to start, because taking action—even poorly—makes the second step much easier.
• Silence the inner critic: Ignore self-doubt and procrastination by forcing yourself to get your thoughts onto paper or a whiteboard immediately.
• Spark collaboration: Once an idea is out in the open, even a flawed one, it acts as a catalyst for the group to naturally revise, refine, and build upon it together.

I forked github.com/gijsdev/ublock-hide-yt-shorts, did some cleanup and added a few more filters to hide all YouTube Shorts.
I will maintain this fork in the future.

uBlock filter list to hide all YouTube Shorts

I stumbled upon this quite bonkers article about building a web search engine from scratch as a solo developer with relatively modest resources, i absolutely can recommend reading it:

Building a web search engine from scratch in two months with 3 billion neural embeddings

Summary (Generated):

• Wilson Lin built a full web search engine from scratch in ~2 months, crawling ~280M pages and generating 3B SBERT embeddings on a GPU cluster consisting of 200 GPUs.
• Lin focused on neural-embedding search, with smart HTML normalization + sentence-level chunking + contextual “statement chaining” so queries match meaning and intent, not keywords.
• The infra is highly optimized + cheap: custom crawler & RocksDB-based queues/KV, sharded HNSW / CoreNN vector DB, mTLS service mesh, hundreds of GPUs on low-cost providers (Runpod, Hetzner, Oracle).
• The SERP emphasizes high-quality, low-SEO-spam content, knowledge panels (Wikipedia/Wikidata), and a light AI assistant for quick answers and reranking, but still feels like a classic fast search engine.
• Biggest lessons: crawling + quality filtering are the hardest part, embeddings make very specific queries vastly better, and search + LLMs will likely coexist (LLMs shouldn’t memorize everything, but retrieve via dense indices).

This is a super cool leaderboard for lossless text compression via NLP (and yes, that includes AI)! The top solution manages to compress the first GB of the English Wikipedia to a whopping 10.7% of its original size, including the compression program itself!

Hyper Text Compression: Shrinking Wikipedia to 10.7% of its Size!

Automatic TLDR by Gemini 2.5 Pro:

This page describes the Large Text Compression Benchmark, an open competition that ranks lossless data compression programs. The primary goal is to encourage research in artificial intelligence (AI) and natural language processing (NLP) by treating text compression as a language modeling problem.

Benchmark Overview

• Test Data: The benchmark uses the first $10^9$ bytes (1 GB) of an English Wikipedia XML dump from March 3, 2006, known as enwik9.
• Ranking Metric: Programs are ranked solely by the total size, which is the sum of the compressed enwik9 file size and the size of the zipped decompresser program. A smaller total size is better.
• Secondary Information: Data such as compression/decompression speed and memory usage are provided for informational purposes but do not influence the rankings.
• Goal: The benchmark’s main purpose is not to find the best general-purpose compressor but to push the boundaries of data modeling, a fundamental challenge in both AI and compression.

Key Findings and Algorithms

The results table shows a wide variety of compression programs, ranked from the best compression ratio to the worst. A clear trend emerges from the top-performing entries:

• Dominance of AI Models: The highest-ranking compressors, such as nncp and cmix, utilize sophisticated AI-based algorithms. These include neural network models like Transformers (Tr) and Long Short-Term Memory (LSTM), as well as advanced Context Mixing (CM) techniques. These methods excel at modeling the complex patterns in natural language text, resulting in superior compression ratios.
• Trade-offs: There is a significant trade-off between compression ratio, speed, and memory. The top-ranked AI-driven compressors are extremely slow and require vast amounts of memory (often many gigabytes) and, in some cases, specialized hardware like GPUs.
• Traditional Algorithms: More conventional algorithms like Lempel-Ziv (LZ), Burrows-Wheeler Transform (BWT), and Prediction by Partial Match (PPM) are found further down the list. While they are generally much faster and use less memory, they cannot achieve the same level of compression as the leading AI models on this specific text-based task.

Hutter Prize

The benchmark is closely related to the Hutter Prize, which offers prize money for open-source compression improvements on a smaller subset of the data (enwik8, the first $10^8$ bytes). This prize has specific hardware and time constraints, encouraging practical advancements in the field.

This is a article that explores the capabilities of OpenAI’s o3 model. HN

The impacts of AI reasoning are getting closer and closer to surpassing human capabilities.
But running it to solve a problem is extremely expensive.

Effectively, o3 represents a form of deep learning-guided program search. The model does test-time search over a space of “programs” (in this case, natural language programs – the space of CoTs that describe the steps to solve the task at hand), guided by a deep learning prior (the base LLM). The reason why solving a single ARC-AGI task can end up taking up tens of millions of tokens and cost thousands of dollars is because this search process has to explore an enormous number of paths through program space – including backtracking.

(programs is defined here as “My mental model for LLMs is that they work as a repository of vector programs”)

As far as i understand this and it’s implications i appears to me this process is more like a directed brute force with many candidates and selects the best one.
There is a thread on HN that discusses this topic.

GPT-4 Summary

OpenAI’s new o3 system has achieved a groundbreaking 75.7% on the ARC-AGI-Pub Semi-Private Evaluation within a $10k compute limit, surpassing previous models. A high-compute version scored 87.5%, marking a major step in AI’s adaptability to novel tasks. This success highlights a shift from scaling existing architectures to innovative mechanisms for generalization and test-time knowledge recombination.

Key achievements:

• ARC-AGI performance: o3 shows unprecedented adaptability, unlike previous GPT-family models, which struggled with novel tasks.
• Efficiency challenges: Low-compute costs $17–$20 per task but high-compute performance remains expensive and exploratory.
• Core innovation: o3 employs natural language program search, generating and executing task-specific solutions during runtime, guided by deep learning priors.

Despite its advancements, o3 is not AGI, as it still fails simple tasks. Upcoming benchmarks, such as ARC-AGI-2 in 2025, aim to further challenge AI systems while encouraging open-source progress.

This breakthrough underscores a qualitative leap in AI research, reshaping paths toward AGI and sparking broader scientific engagement.

I stumbled on HN over this very interesting article about a new kind of context memory system that, is able to remove information that is “unhelpful or redundant details”.

Thinking further, i think this would be super helpful for semantic search, that is currently not very performant due to the missing filters that extract importance. I have tried to counter this problem until now via summarization through small LLMs, but as one might guess turns out as not very precise and super expensive. There are other ideas one could post process text with LLMs but they are not very efficient either.

Paper https://arxiv.org/abs/2410.13166

TLDR by GPT-4o

(the article is very good, you might to prefer to read it over this TLDR, but here is it anyway)

Sakana AI introduces Neural Attention Memory Models (NAMMs), a novel memory system for transformers inspired by human selective memory. NAMMs optimize how transformers store and retrieve information, enabling them to “remember” important tokens and “forget” redundant ones, significantly improving efficiency and performance, particularly for long-context tasks.

Key Technical Highlights:

1. Evolutionary Optimization: NAMMs use evolutionary algorithms to train a neural classifier that decides which tokens to keep or discard, bypassing non-differentiable challenges.
2. Execution Steps:
   • Convert attention sequences into spectrograms.
   • Compress data using an exponential moving average (EMA).
   • Use a classifier to score and selectively prune tokens.
3. Generalization:
   • NAMMs can zero-shot transfer to other transformers (e.g., vision, reinforcement learning) without retraining.
   • They adapt to tasks differently—retaining global information in early layers and focusing on local details in later ones.
4. Performance: Tested on LongBench, InfiniteBench, and their Japanese benchmark ChouBun, NAMMs reduce memory usage and outperform prior hand-designed memory strategies (H₂O, L₂).

NAMMs demonstrate cross-domain mastery and efficiency gains across diverse tasks and input modalities, paving the way for future research into learning transformers directly atop evolved memory systems.

I want to share this great article about generic coding in Go: https://bitfieldconsulting.com/posts/constraints

Also Go Proverbs is quite nice!

Ohh and this Blog has now footnotes and popups for footnotes! ¹

Summary (Generated):

In this tutorial, John Arundel explores constraints in Go, the paradoxical feature of generics that simultaneously limits and expands possibilities. By delving into method sets, type elements, unions, intersections, and approximations, he demonstrates how constraints enable precise operations on type parameters. Whether you’re learning to leverage fmt.Stringer, creating flexible numeric constraints, or handling derived types, this guide unlocks a deeper understanding of Go’s generics—perfect for honing your skills or tackling practical challenges in coding.

Opportunities for AI in Accessibility

Personal Note:

I think this article falls short in capturing the full impact AI will have on accessibility, not only for people with disabilities but also for those who, despite not having recognized disabilities, face accessibility issues. This includes tasks ranging from reading cursive handwriting to understanding non-native speakers who use dialects.

Moreover, I believe there are numerous additional areas of accessibility that the article didn’t address or mention. For example, the sound recognition feature of iOS saves lives by alerting users to critical sounds they can’t hear, like a smoke detector alarm. Likewise, large language models (LLMs) are incredibly useful for people with ADHD (this was indirectly mentioned int the article (props)) by transforming texts into bullet points or adding better text breaks for improved readability. People with dyslexia also benefit significantly from LLMs like GPT-4, which can correct typos and grammar errors in ways far superior to any non-AI solution.

There are many more examples like these, and we’re only scratching the surface of what’s possible and how AI is being used to make the world more accessible in unexpected ways.

TLDR:

1. The author acknowledges Joe Dolson’s skepticism towards AI, particularly in the realm of accessibility, but aims to highlight where AI can positively impact people with disabilities (PwD) without disputing Dolson’s concerns.
2. The discussion includes the potential for AI in generating alternative text for images, emphasizing the need for improvement in contextual understanding and human-in-the-loop systems for better alt text authoring.
3. The article suggests AI’s future capabilities could revolutionize access for PwD by enabling interactive queries on images, simplifying complex charts for better understanding, and even converting visual data into more accessible formats.
4. It addresses the significant issue of bias in algorithms, proposing that increased diversity in algorithm development could mitigate harm and benefit PwD, with examples like the employment platform Mentra designed for neurodivergent individuals.
5. The author argues for the importance of diverse teams and data in creating more inclusive AI systems, underscoring the potential of AI to empower PwD while acknowledging the ongoing risks and advocating for responsible development and use.

Book Don’t Die

Personal Note:

Since Bryan has started selling food replacement products aimed at promoting longevity, I am inclined to interpret his statements as being influenced by his interest in selling these products. In my view, all food replacement products are scams, and none can match the health benefits of a whole food, plant-based diet.

Despite this, I believe the book offers valuable insights. I’m open to opportunities for personal growth, so I’m curious to understand his perspective and see if it contributes to my own development.

TLDR:

The book “DON’T DIE_by Zero” by Bryan Johnson seems to focus on a few major themes and actionable goals, as evident from the selected excerpts:

1. Self-Improvement Through Science: The author emphasizes the importance of not accepting the status quo, especially regarding health and well-being. This is evident in his approach to combatting the effects of aging and depression through scientific means, such as tracking biomarkers and experimenting with gene therapy.

2. Reassessment of Basic Assumptions: The book encourages questioning basic life assumptions and societal norms. It promotes a scientific reassessment of everyday queries like sleep, diet, and health, moving beyond traditional folk wisdom to a more data-driven approach.

3. Embracing Technological Advancements: The author discusses humanity’s potential at an “unprecedented inflection point” due to technological advancements. He suggests that these advancements could allow us to reprogram the basic building blocks of our environments and ourselves, indicating a belief in the potential for self-directed evolution.

4. The Importance of Purpose and Facing Mortality: A key moment in the book is when the narrator, aware of his impending death, asks his friends what they would do if they were dying. This reflects the theme of confronting mortality and the importance of finding purpose and meaning in one’s actions.

In summary, the actionable goals extracted from this book are:

• Utilize scientific and technological advancements for personal health and self-improvement.
• Reassess and question traditional beliefs and norms in light of modern scientific understanding.
• Embrace the potential of technological evolution to improve humanity’s future.
• Reflect on one’s mortality to find deeper meaning and purpose in life.

The book appears to blend personal narrative with a call for a broader societal and technological revolution, focusing on self-improvement, scientific inquiry, and the significant potential of technological advancements.

The quality of your life is the quality of the people you get to know: Illuminating the David Brooks way

TLDR:

The article discusses the importance of social connections and interpersonal skills, aligning with the ideas presented in David Brooks’ book “How to Know a Person: The Art of Seeing Others Deeply and Being Deeply Seen.” The main argument is that life’s purpose revolves significantly around our relationships with others. It emphasizes that modern society often neglects teaching essential social skills, like active listening, showing empathy, and effective communication, leading to increased loneliness and disconnection.

Key insights include:

1. Social Skills Deficiency: The lack of training in vital social skills in educational systems contributes to a disconnect in society.
2. Impact of Technology: Technology, while making solitude more bearable, also hinders the development of meaningful connections.
3. Importance of Deep Conversations: Engaging in profound discussions is crucial for meaningful relationships, yet modern society often avoids such interactions.
4. Brooks’ Typology: The concept of “Diminishers” and “Illuminators” is introduced, where Diminishers make people feel unseen and used, while Illuminators actively work to understand and uplift others.
5. Practical Application: The article suggests that actively working on social skills can improve one’s ability to connect deeply with others.

Actionable takeaways:

• Develop Interpersonal Skills: Focus on enhancing skills like empathy, active listening, and clear communication.
• Engage in Meaningful Conversations: Seek and initiate deeper conversations beyond superficial topics.
• Be Aware of Social Dynamics: Understand the difference between Diminishers and Illuminators in social interactions and strive to be an Illuminator.
• Utilize Relationships for Growth: Use relationships as a means to understand oneself better and grow personally.

The article intertwines personal experiences and observations with the book’s concepts, providing a comprehensive view of the importance of interpersonal skills and deep connections in leading a fulfilling life.

The Evolution Of What It Means To Be Human

TLDR:

The article “The Evolution Of What It Means To Be Human” by Nathan Gardels, editor-in-chief of Noema Magazine, explores the concept of human identity in the context of our interactions with the natural cosmos and the environment we have formed through technology and culture. The author argues that what it means to be human is not a constant, but is continually constituted, altered, and re-constituted through our interface with an open and evolving world.

The article features insights from Benjamin Bratton, a philosopher of technology, who suggests that technology and science do not erase ’the Human’ but rather reveal what humans have always been. He argues that instead of leading to a ‘post-human’ condition, science and technology might lead to a slightly more human condition. Bratton suggests that our understanding of ourselves is enhanced, not diminished, by scientific and technological advancements.

Bratton also discusses the idea of anthropo-technogenesis, the process by which humans and technology evolve together. He suggests that this process should not be seen as a threat to human agency or autonomy. Instead, it reveals the nature of subjectivity and the qualities of being a human body. He argues that we are not separate from the world, but are embedded in it, and this shapes our identity.

The article concludes by discussing the implications of these ideas for politics and governance. It suggests that we need a theory of governance that recognizes the promise of scientific and technological advances while affirming the autonomy of humans, albeit reconfigured through a new awareness. This would involve a balance between the capacity to act and the ability to amend or arrest action that could lead to harmful consequences.

Book summary: 2k to 10k

TLDR:

“2k to 10k: Writing Faster, Writing Better, and Writing More of What You Love” by Rachel Aaron provides practical tips and advice for writers. It’s divided into two parts: writing process and background work.

Part one discusses a process for writing quickly and efficiently, which is based on three factors: knowledge, time, and enthusiasm. It encourages understanding what you’re going to write before you start, optimizing your writing sessions based on when/where you’re most productive, and writing content that you’re enthusiastic about.

Part two shares advice on plotting, character development, and editing. It provides a five-step process for plotting a story and emphasizes the importance of well-developed characters with agency. It also breaks down the traditional three-act structure, the concept of scenes serving multiple purposes, and provides a practical approach to editing.

The book underlines that writing is an enjoyable process, characters should be designed to drive their stories, and editing is a skill that can be mastered. It encourages authors to shift their thinking and approach to writing, making it both an enjoyable and efficient process. The final objective of editing, according to the book, is to enhance the reader experience.

What Should You Do with Your Life? Directions and Advice

TLDR:

The article “What Should You Do with Your Life? Directions and Advice” by Alexey Guzey provides guidance for people who are unsure about their life plans. Here’s a summary:

1. What to work on: Guzey suggests several resources for finding interesting problems to work on, including Y Combinator’s Requests for Startups, OpenAI’s Requests for Research, and various lists of research ideas and project ideas.

2. How to work on the problem you like: Guzey advises writing down a shortlist of exciting problems, selecting the most exciting and tractable ones, and then learning everything you need to start working on the problem. This could involve taking MOOCs, auditing classes at a local university, or finding a tutor. Once you’ve gained sufficient knowledge, start working on a project related to the problem, produce a minimal public impressive result, and share it with people working in your area.

3. Cold emails and Twitter: Guzey emphasizes the effectiveness of cold emails and Twitter for connecting with people who can help you. He shares examples of his own successful cold emails and provides tips for writing them.

4. Finding funding: For projects that require funding, Guzey lists several resources, including Nadia Eghbal’s list of microgrants, the Thiel Fellowship, and Emergent Ventures.

5. General advice: Guzey shares advice from various sources, such as Paul Graham’s essay “How to Do What You Love”, Patrick Collison’s advice for ages 10-20, and Patrick McKenzie’s “Don’t End The Week With Nothing”. The common theme is to be proactive, produce tangible work, and take advantage of opportunities.

6. More projects to work on: Guzey ends the article by suggesting more projects to work on, like Jacob Trefethen’s Vigilante accountability project.

A Project of One’s Own

TLDR:

The article, titled “Own”, is written by Paul Graham and was published in June 2021. It discusses the concept of working on a project of one’s own and how it differs from ordinary work. Here are the key points:

1. Working on a project of one’s own is compared to skating, which is more fun and productive than ordinary work (walking). This kind of work is often exciting and engaging, even though it may not always lead to happiness due to the challenges and problems encountered.

2. The author argues that a significant proportion of great work has been done by people who were “skating” in this sense. He emphasizes the importance of autonomy and the sense of ownership in work.

3. The article also discusses the disconnect between the work children do (like building a treehouse) and the work adults do. It criticizes the school system for not making it clear to children that their playful projects could lead to serious work in adulthood.

4. The author prefers his children to work on ambitious projects of their own rather than focusing on getting good grades. He argues that such projects have more predictive value for future success.

5. The article also explores the concept of collaboration and how it fits into the idea of owning a project. It suggests that collaboration can either involve sharing a single project or working on separate projects that fit together.

6. The author concludes by encouraging adults to recapture the carefree confidence they had as children when starting new projects. He suggests that adults can have the best of both worlds: being deliberate in choosing to work on projects of their own and carelessly confident in starting new ones.

Ask HN: Favorite Blogs by Individuals?

TLDR:

The discussion on the provided Hacker News link revolves around various topics, primarily focusing on the appreciation of a historian’s blog that frequently appears on Hacker News. The blog is praised for its detailed and technical descriptions of mechanical processes, which some users liken to the work of a programmer. The author’s approach to discussing topics such as archery, which includes elements of physics, empirical studies, historical anecdotes, and comparisons to modern video games and TV, is also appreciated.

There’s also a discussion about a blog post on medieval medicine, which is praised for its insightful exploration of the topic. Some users express their appreciation for the writer’s style, which includes the use of parentheses, italics, and bolding to clarify the structure of thoughts.

The thread also includes a discussion about a blog called “A half-hour to learn Rust,” which is recommended for those interested in learning Rust with a good programming background but no specific Rust knowledge.

There are also mentions of other blogs and resources, such as diff.blog, a platform for discovering new developer blogs, and a discussion about the cost of domain names. Some users share their favorite blogs and discuss the benefits of using an RSS feed reader.