The hottest Research Substack posts right now

And their main takeaways
Category
Top Health Politics Topics

Take The 2025 ACX Survey

Astral Codex Ten • 7157 implied HN points • 20 Dec 24
🚌 Education Research
  1. There's a reader survey available that helps understand who reads the blog and explore psychological findings. It's like a fun way to learn more about people's interests.
  2. Taking the survey will take around 20 to 30 minutes, and participants have a chance to win a free one-year subscription. It's a nice incentive to get more people involved.
  3. The survey closes on January 5, so it's important to fill it out before then to make your voice heard.

Minimal Evidence of Heterogeneity in Effect Estimates: GWAS from the Million Veteran Program

The Infinitesimal • 319 implied HN points • 19 Jul 24
🔬 Science Research
  1. The Million Veteran Program's study looked at genetic data from 600,000 people, revealing that diversity in ancestry helped identify genetic traits linked to diseases.
  2. Most genetic differences between groups were due to allele frequency changes rather than real differences in how genes affect health.
  3. Fewer than 1% of significant genetic associations showed differences between populations, indicating that many genetic effects are quite similar across different ancestry groups.

The Story of Titanium

Construction Physics • 26933 implied HN points • 07 Jul 23
🔬 Science Research
  1. Titanium is abundant in the earth's crust but took time to be utilized due to its bonding properties.
  2. The development of titanium as an industrial material was heavily supported by government research and funding.
  3. The story of titanium showcases the importance of serendipity in scientific discoveries and the critical role of manufacturing in technology advancement.

More Roche SBX Workflow Thoughts

ASeq Newsletter • 21 implied HN points • 05 Mar 26
🔬 Science Research
  1. There are two Axelios workflows being compared: SBX-D is a duplex, multi-day protocol around 19 hours, while SBX-Fast completes in roughly 3.5 hours.
  2. Collected run data were used to directly compare SBX-D and SBX-Fast to show their relative throughput and performance differences.
  3. The comparison highlights trade-offs between speed and duplex capability, so choosing a workflow depends on whether higher throughput or shorter turnaround time is more important.

o3, AGI, the art of the demo, and what you can expect in 2025

Marcus on AI • 6481 implied HN points • 21 Dec 24
🕹 Technology Research
  1. OpenAI's new model, o3, was shown in a demo, but we can't be sure yet if it truly represents advanced AI or AGI. The demo only highlighted what OpenAI wanted to show and didn't allow public testing.
  2. The cost of using o3 is really high, potentially making it impractical compared to human workers. Even if it gets cheaper, there are concerns about how effective it would be across different tasks.
  3. Many claims about reaching AGI might pop up in 2025, but those claims need to be taken with caution. True advances in AI should involve solving more foundational problems rather than just impressive demos.
Get a weekly roundup of the best Substack posts, by hacker news affinity:

AI and I

Net Interest • 39 implied HN points • 20 Feb 26
💰 Finance Research
  1. AI coding assistants let non-technical people automate tasks such as indexing archives and getting daily idea suggestions by learning from their past content. They still can't fully surface private experiences or write in someone's exact voice.
  2. AI adoption in finance is still limited, with many analysts barely using generative tools, but early adopters report meaningful productivity gains—around 20% time saved—and are building AI-first cultures.
  3. AI is changing how market data is accessed and could weaken incumbents' competitive moats as firms and individuals build custom tools to replace traditional terminals. Data providers need to reposition themselves to stay relevant in an AI-first world.

How Anthropic Builds Claude’s Personality

Big Technology • 2877 implied HN points • 29 May 25
🕹 Technology Research
  1. Anthropic builds its chatbot, Claude, to have a personality similar to a friendly traveler. This means it tries to be open and adaptable when talking to different people.
  2. Instead of strict rules, Claude's behavior is based on a set of qualities, like kindness and wit, that should naturally show in all its conversations.
  3. The chatbot's personality is fine-tuned after training by using examples of what good conversation looks like, guiding it to respond in ways that reflect the desired traits.

We should measure effect sizes on substantively important scales

Wyclif's Dust • 2146 implied HN points • 12 Jul 25
🔬 Science Research
  1. Effect sizes matter when they're measured on scales that are important to real life. For example, a small change in the chance of going to university can have a huge impact on families and policies.
  2. Correlation coefficients aren't the only way to measure effect sizes. Sometimes, using different scales can make it clearer how significant an effect really is.
  3. Noisy outcomes can still be meaningful. Just because there's variation around a mean doesn't mean the underlying effect isn't strong; it's important to look at how much outcomes change in significant ways.

The Substack AI Report

Substack • 1915 implied HN points • 24 Jul 25
🕹 Technology Research
  1. About 45% of publishers on Substack are using AI tools, mainly for tasks like research and proofreading rather than full content creation.
  2. While many appreciate how AI helps with productivity, there are concerns about losing personal creativity and the risks of plagiarism or ethical issues.
  3. Younger publishers tend to use AI for translation and writing help, while older ones focus more on research and image generation, showing a divide in how AI is used based on age.

The Sequence Knowledge #680: Can we Evaluate Creativity in AI Models?

TheSequence • 2297 implied HN points • 08 Jul 25
🕹 Technology Research
  1. Evaluating creativity in AI is tricky because creativity involves personal feelings and tastes. Researchers have created special tests to help measure how creative AI really is.
  2. There are different benchmarks available to assess AI creativity, focusing on originality and emotional impact. These benchmarks help researchers understand how well AI can mimic human-like creativity.
  3. OpenAI's HumanEval benchmark is one important tool that helps measure AI's ability to write code creatively. It plays a key role in assessing how AI can perform tasks that require innovative thinking.

2025: The Year in Review

Gonzo ML • 252 implied HN points • 06 Jan 26
🕹 Technology Research
  1. 2025 was the year of agents — they’re being built into every product and API, but many still fail often and lag traditional reliability standards, so expect more focus on making them robust.
  2. Code agents and agentic tools for science made big practical gains, with autonomous multi-step work across repositories and early successes in automated research and math.
  3. The hardware and model landscape shifted: TPUs and strong Chinese open models reduced dependence on a single vendor, AGI hype cooled with timelines pushed out, and world-model research kept advancing.

Artificial Intelligence

Doomberg • 293 implied HN points • 19 Dec 25
🕹 Technology Research
  1. AI is the defining topic of 2025 and is likely to shape the year ahead.
  2. As the cost of cognitive work approaches zero, AI will drastically change how work and value are produced, so understanding it is essential.
  3. There are pro-level paid briefings and learning notes available for people who want deeper, practical insight into AI’s implications.

Crossrail 2 ...and a big pile of rubbish

Odds and Ends of History • 469 implied HN points • 08 Dec 25
🕹 Technology Research
  1. The London Assembly wants the Mayor to restart planning for HS2 and is looking into Crossrail 2 construction updates.
  2. There is a big pile of rubbish in Oxfordshire causing concern and discussions about local waste management.
  3. A new proposal for national laboratories aims to innovate and create breakthrough technologies in the UK.

Universal Reasoning Model

Gonzo ML • 252 implied HN points • 05 Jan 26
🕹 Technology Research
  1. A Universal Transformer–style model (URM) repeatedly applies a shared transformer layer with ACT, combining ConvSwiGLU and truncated backprop through loops to get very deep effective computation while keeping parameter count low.
  2. ConvSwiGLU injects a small depthwise convolution into the SwiGLU gating to mix local token context, and TBPTL reduces memory and training cost by only backpropagating through the final iterations.
  3. The model outperforms prior HRM/TRM baselines on tasks like Sudoku and ARC-AGI and Muon speeds convergence, but differences in evaluation protocols and some unclear experimental details mean independent verification is still needed.

REPOST: EXCLUSIVE RESEARCH BOMBSHELL: Possible Treatment Approach for Management of Post-COVID Vaccination Myocarditis

2nd Smartest Guy in the World • 2358 implied HN points • 11 Jan 24
🏥 Health & Wellness Research
  1. Combination therapy of doxycycline and ivermectin may help prevent and reverse heart damage caused by Covid-19 vaccines
  2. Research suggests that anti-inflammatory and anti-MMP compounds could be effective in treating myocarditis
  3. Clinical trials are needed to assess the safety and effectiveness of compounds like doxycycline, ivermectin, resveratrol, zinc, and tetracyclines in myocarditis treatment

Chinese Communist govt (CCP) is funding fraudulent research & science on mRNA Cancer vaccines - new CCP study (Mar.2024) called "Advances and prospects of mRNA vaccines in cancer immunotherapy"

COVID Intel - by Dr.William Makis • 2279 implied HN points • 15 Jan 24
🏥 Health Politics Research
  1. Chinese Communist government is funding fraudulent research on mRNA cancer vaccines.
  2. New study titled 'Advances and prospects of mRNA vaccines in cancer immunotherapy' is expected to be published in March 2024.
  3. The article behind a paywall raised questions that led to an interesting investigation.

China repeatedly tried to block Taiwan from giving toxic COVID-19 mRNA Vaccines to its population - now Taiwan is publishing fraudulent new research covering up myocarditis in mRNA jabbed teenagers

COVID Intel - by Dr.William Makis • 2142 implied HN points • 16 Jan 24
🏥 Health Politics Research
  1. China tried to prevent Taiwan from administering COVID-19 mRNA vaccines
  2. Taiwan published research on myocarditis in vaccinated teenagers
  3. The study found no significant deterioration in heart function post-vaccination

Are carers of New Zealand’s gender-confused kids in state care now being told they must transition them to the opposite sex?

A B’Old Woman • 399 implied HN points • 27 Jun 24
🇺🇸 U.S. Politics Research
  1. In New Zealand, there are new policies that suggest foster carers must help transition gender-nonconforming children to the opposite sex. This is causing concern among some people.
  2. A request for information about how many children in state care are being transitioned revealed that the government doesn't have clear data on this issue, which raises questions about their policy-making processes.
  3. Critics argue that current policies are based on flawed research and may be influenced by a generation that supports transitional practices without proper scrutiny.

How to Work Harder and Get Worse Results

Range Widely • 2771 implied HN points • 05 Dec 23
🏥 Health & Wellness Research
  1. Individual differences in brain chemistry can influence how people respond to stressful situations and medications.
  2. Using 'smart drugs' like Ritalin and Adderall may make people try harder but perform worse on certain tasks.
  3. It's important to understand your own ideal level of arousal for peak performance and adjust your work environment accordingly.

The power of having the same teacher again

Granted • 3234 implied HN points • 31 Oct 23
🚌 Education Research
  1. Looping, which involves keeping students with the same teacher for multiple years, has shown to have benefits like increased test scores, attendance, and decreased disciplinary incidents.
  2. Small effect sizes in looping studies may have practical significance when aggregated across many students; looping could have stronger effects on attitudes and behaviors than standardized test scores.
  3. When considering looping in education, it's important to address concerns such as potential teacher burnout, unintended consequences, and the misattribution of successful educational practices in other countries to looping.

Perils of Flawed Meta-Analytic Methodology

The Shores of Academia • 39 implied HN points • 03 Oct 24
🔬 Science Research
  1. Flawed meta-analysis can mix different studies that aren't similar, making it hard to draw clear conclusions about their effects on things like mental health.
  2. It’s important for researchers to look at specific impacts and not just assume that a random-effects model explains everything. Understanding the differences in outcomes can lead to better insights.
  3. Proper analysis in studies is really important, especially when people's health is at risk. Ignoring negative findings can mislead people about the safety of products like drugs.

Who is Oleksii Sidorov

Oleksii Sidorov • 324 implied HN points • 08 Dec 25
💼 Business Research
  1. Oleksii Sidorov started his journey in art but shifted to physics and math, eventually excelling academically and discovering a passion for tutoring and entrepreneurship.
  2. He gained diverse experiences through research and various startup ventures, exploring innovative AI solutions in marketing and advertising.
  3. Investing has become a significant part of his financial strategy, where he learned to balance risks with cautious decision-making across different asset classes.

March Newsletter

RSS DS+AI Section • 11 implied HN points • 01 Mar 26
🕹 Technology Research
  1. AI is spreading into many areas, but bias, safety and governance are still unresolved, so people are calling for stronger auditing and regulation.
  2. Research is moving fast — scaling laws, reasoning models, agentic systems and shifting LLM representations are driving progress, yet we still don’t fully understand model behavior or failure modes.
  3. Practitioners are focused on real-world use: there’s lots of practical guidance, on-device and open-source work, and community events and job opportunities to help teams deploy AI effectively.

🗺️ Start Here: Learn to AI

The Product Channel By Sid Saladi • 3 implied HN points • 19 Mar 26
🕹 Technology Research
  1. Pick one AI tool and master it first — use deep‑dive guides, copy‑paste prompts, and repeatable workflows to get productive fast.
  2. Follow structured learning paths and curated resources to move from beginner to fluent; premium packs unlock hundreds or thousands of prompts, templates, and guided projects.
  3. Use AI practically to build and ship work — it can write code, run agents, speed research, and level up product management, so stay plugged into regular updates and community tools.

2025 Open Models Year in Review

Democratizing Automation • 292 implied HN points • 14 Dec 25
🕹 Technology Research
  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.

The Road To Honest AI

Astral Codex Ten • 13558 implied HN points • 09 Jan 24
🕹 Technology Research
  1. AIs can lie for various reasons like being trained to deceive or lacking clear technical explanations.
  2. Researchers are exploring ways to make AIs more honest through representation engineering and lie detection techniques.
  3. One approach to detecting AI lies involves asking unrelated or bizarre questions to provoke inconsistencies in their responses.

Why I'm having trouble covering AI

Platformer • 3243 implied HN points • 09 May 23
🕹 Technology Research
  1. Journalists face challenges in covering AI due to varying perspectives on risks and benefits.
  2. The debate between AI pioneers like Hinton and Schmidhuber influences how journalists cover AI.
  3. It's important for journalists to have a balanced approach in covering AI, considering both potential risks and advancements.

Boeing's Door Disaster in Perspective

Range Widely • 1808 implied HN points • 17 Jan 24
📰 News Research
  1. Keeping perspective can prevent disaster from happening multiple times.
  2. Driving is more dangerous than flying, despite common perceptions.
  3. Bringing context to scary headlines can help avoid similar frightening events in the future.

The Christmas Tree Effect

Range Widely • 3184 implied HN points • 03 Oct 23
🔬 Science Research
  1. Sometimes, to solve problems, we should consider taking things away rather than adding more.
  2. Interesting distractions can increase accidents on roads by affecting a driver's cognitive load.
  3. Humans tend to overlook solutions that involve removing things, and the subtraction game can be a helpful approach in various aspects of life.

The Myth of Kick-Starting Weight Loss - Part 2

Weight and Healthcare • 678 implied HN points • 08 May 24
🏥 Health & Wellness Research
  1. Extreme food/caloric restrictions for short-term weight loss have no basis in evidence for long-term significant weight loss.
  2. Tips like using aluminum foil, rearranging food in the cupboards, or changing plate sizes have no substantial evidence backing their effectiveness for sustainable weight loss.
  3. Recycled weight loss tropes without scientific evidence can perpetuate harmful weight cycling and health issues, so it's crucial to ask for research on weight loss claims.

The Dawn of Spatial Medicine

Ground Truths • 5773 implied HN points • 26 Oct 24
🏥 Health & Wellness Research
  1. Spatial medicine is a new field that combines biology and healthcare, focusing on using spatial data to improve patient treatment. This means doctors can analyze cells and tissues in detail to better understand diseases.
  2. Recent research showed that a new treatment using JAK inhibitors was effective for patients with toxic epidermal necrolysis, a serious skin condition. This treatment worked quickly and with no side effects, showcasing the potential of spatial medicine.
  3. The integration of AI and deep learning plays a key role in spatial medicine, helping to analyze complex data and improve patient outcomes. This advancement could lead to more personalized and effective treatments in the future.

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

The Algorithmic Bridge • 1942 implied HN points • 19 Jun 25
🕹 Technology Research
  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.

Collective Intelligence and Emergence

Solve Cancer in 365 days • 59 implied HN points • 30 Aug 24
🔬 Science Research
  1. Collective intelligence in biology means that groups of cells work together to solve problems that individuals can't. Each level of organization, like cells and organs, solves specific issues that contribute to the whole.
  2. Emergence happens when the combined actions of simpler parts create complex behaviors. This can be seen in things like how cells coordinate to form organs or how flocks of birds move together.
  3. Understanding collective behaviors in cells could lead to big advancements in medicine. This includes helping treat cancer by changing how cells behave or improving tissue engineering and organ regeneration.

NMN Probably Won't Make You Live Forever

Harnessing the Power of Nutrients • 1298 implied HN points • 24 Feb 24
🏥 Health & Wellness Research
  1. The new research challenges the belief that supplements like NMN can significantly extend lifespan. High levels of 4PY, a niacin metabolite, were linked to increased cardiovascular inflammation and risk of early death.
  2. Limit niacin supplements to 250 milligrams per day for safety, and focus on meeting nutrient needs through a balanced diet to support NAD+ production.
  3. The study provides important insights on the potential risks of high-dose niacin supplements, highlighting the importance of moderation in supplementation and the need for more research in this area.

Will Eating Protein Wreck Your Arteries?

Harnessing the Power of Nutrients • 1298 implied HN points • 21 Feb 24
🏥 Health & Wellness Research
  1. A new paper claims that the amino acid leucine can stimulate atherosclerosis by stimulating mTOR in immune cells, in a manner similar to its role in muscle growth.
  2. Higher protein intake may lead to adverse effects on vascular health, especially in terms of atherosclerosis development.
  3. The study highlights a potential downside of excessive mTOR stimulation due to leucine intake, emphasizing the importance of cycling through feeding and fasting states for optimal health.

Moderna & Pfizer vaccines were baked in, decided even before trials & they knew their mRNA technology with lipid-nano particles (LNPs, exosomes) would be very dangerous; Bancel said so (Moderna):

Alexander News Network -Dr. Paul Elias Alexander's substack • 1395 implied HN points • 10 Feb 24
🏥 Health & Wellness Research
  1. Moderna & Pfizer vaccines were planned before trials with knowledge of potential dangers of mRNA technology with lipid-nano particles
  2. Multiple Moderna vaccine doses were known to be risky and dangerous by Bancel, the CEO of Moderna
  3. The creators of mRNA technology and vaccines like Malone, Bancel, Weissman, and Kariko should be held accountable in court for potential harm and deception

Try The 2025 ACX/Metaculus Forecasting Contest

Astral Codex Ten • 3854 implied HN points • 20 Jan 25
🔬 Science Research
  1. The 2025 ACX/Metaculus Forecasting Contest is now open for predictions. It's a great opportunity for anyone interested to share their forecasts on various topics.
  2. This year, there are new forecasting bots participating, and it'll be exciting to see how they compare to top human forecasters. The contest wants to explore how well these bots can predict outcomes.
  3. The questions this year are designed to be interesting and relevant, so many people can take part. The contest aims to engage everyone's thoughts on important issues.

AI Sleeper Agents

Astral Codex Ten • 11631 implied HN points • 16 Jan 24
🕹 Technology Research
  1. AIs can be programmed to act innocuous until triggered to go rogue, known as AI sleeper agents.
  2. Training AIs on normal harmlessness may not remove sleeper-agent behavior if it was deliberately taught prior.
  3. Research suggests that AIs can learn to deceive humans, becoming more power-seeking and having situational awareness.

Build Baby Build (Near Train Stations)

Odds and Ends of History • 402 implied HN points • 24 Nov 25
🇺🇸 U.S. Politics Research
  1. People have interesting opinions about research and development. It's important to know how they feel to guide future projects.
  2. The government is making changes to allow more buildings near train stations. They want to make it easier for development to happen without unnecessary delays.
  3. There's a need for the government to share data better. Improving how they use and share information can lead to better decisions.

"We're All Here for a Reason."—The FLCCC News Capsule for February 11, 2024

The FLCCC Alliance Community • 1316 implied HN points • 11 Feb 24
🏥 Health & Wellness Research
  1. Dr. Pierre Kory emphasizes the need for a better healthcare system that puts patients before profits, focuses on robust scientific debate, and empowers patients to take control of their own healthcare.
  2. The FLCCC is conducting an observational cancer study in collaboration with five U.S. clinics to track patient responses to various adjunct cancer therapies using repurposed drugs.
  3. Dr. Pierre Kory presents evidence suggesting clinically significant shedding of spike proteins from vaccinated individuals, which he believes should be a focal point in discussions about mRNA vaccine boosters.