Bookmarks

In 2025, I became a social butterfly the likes of which i’ve never been in my life.

This development coincided with me becoming truly confident, happy, and (mostly) stress-free for really the first time in my life. Huh, how strange.

Introversion, or being anti-social, is just a reflection of how one feels about themself. If you cannot happily, confidently present yourself to the outside world, you will always find it difficult to be social, to be fun around others, and to do so with a carefree, magnetic attitude that draws people in.

It all comes down to what’s in your head.

FRAMEWORK FOR BUILDING A $60K MRR MOBILE APP LIKE THIS

1. Start with a behavior people already have.

Sleeping already happens every night, so the snoring app fits naturally into an existing routine!

2. Tie the app to a question people already think about.

“Do I snore?” often comes up through partners, jokes, or health concerns, so curiosity exists before the app.

This makes your life easier.

3. Use AI to interpret the signal for the user.

The app turns raw audio into meaning by identifying when snoring happens and surfacing what matters.

Tons of niches/ideas are opened up to BECAUSE of AI right now.

4. Deliver the answer immediately.

You wake up, open the app, and see the result from the night before.

Builds the daily habit.

5. Make the result easy to show.

Short audio clips and simple visuals turn into five-second videos, which makes organic sharing effortless.

Virality goes up. Helps with CAC and compounding.

6. Keep the scope narrow and intentional.

The app focuses on one problem instead of becoming a full sleep platform, which keeps it useful and easy to understand.

And the name reflects it too. Says what it does. Autosnore. Snoring recorder.

Perfect.

7. Charge for clarity and peace of mind.

People happily pay to remove uncertainty, especially around health and relationships.

This pattern shows up across a lot of categories because these moments exist everywhere. @ideabrowser has shared a bunch of these ideas.

If people already do the behavior, already ask the question, and react emotionally to the answer, there’s probably a product there.

The fastest way to test it is simple: ship a rough version, show the output on video, and see if people lean in. The response usually tells you quickly whether it’s worth going deeper.

There are more "dumb" categories like this than people realize

Because of LLMs, the only type of people I’d hire now are what I’d call “super-unicorns”

• Developer
• Designer (or great design taste)
• Product-focused (vs feature-focused)
• Marketer
• Agent expert
• High emotional intelligence
• 5+ years experience creating web app products
• Extreme ownership (h/t @jockowillink)

There are very few folks with all these traits …

but I just wouldn’t want to work with someone who didn’t have all these skills/behaviors.

This will keep team sizes very small.

2026 is going to be very interesting for the whole industry.

I’m excited to share a new repo: Agent Skills for Context Engineering

Instead of just offering a library of black-box tools, it acts as a "Meta-Agent" knowledge base. It provides a standard set of skills, written in markdown and code, that you can feed to an agent so it understands how to manage its own cognitive resources.

github.com/muratcankoylan…

Most agent failures are not model failures; they are context failures. This is still an experimental project. The goal is to establish a platform-agnostic standard for context engineering that can be used in Cursor, Claude Code, Copilot or Codex.

skills/
context-fundamentals: What context is, why it matters
context-degradation: How context fails (lost-in-middle, poisoning)
context-optimization: Compaction, masking, caching
multi-agent-patterns: Orchestrator, swarm, hierarchical
memory-systems: Vector RAG, knowledge graphs, Zep
tool-design: Building tools agents can use
evaluation: Testing and measuring agent systems

I believe this is a good start, showing developers how to approach context engineering rather than relying on ready-made tools.

You will also find the aggregated research documents I used to build these skills in the repo. The skills are synthesized from technical blogs on context and prompt engineering that I bookmarked, AI Labs' documentations, and Anthropic Skills examples.

Try the 7 Skills, created using Antrhopic's Skills template format. Experiment with the provided scripts and references, and feel free to contribute to the repo.

Hot take from @filippkowalski - should you remove your paywall from onboarding? The answer might surprise you 👇

Disclaimer: Since we don't really know where the paywall is on a per-app, we use d1 paywall rate as a proxy.

Here's the data:

• n > 1000 (over 1000 apps)
• 30d net arpu (net revenue per user after refunds, for users who installed 45 days ago to 15 days ago, to clear typical trial lengths)
• at least 385 paywalled users
• only looking at Superwall generated revenue
• removed 5% extremes on both ends (top 50 ARPU apps and bottom 50 ARPU apps removed)

Scatter plot - each data point
Histogram on 10% bins

So there is some truth to what Filip is saying - too high a paywall rate may turn off customers before they get to experience your app.

On the flip side, if less than 50% of people are seeing your paywall on day 1, there is almost no shot you'll have a high ARPU.

So while we can't really answer if onboarding is the right place for it, there is definitely such thing as being "too aggressive" with the paywall.

Personally, I love the way @photoroom_app does it - paywall in onboarding, but only after value is demonstrated.

Photoroom guides the user through removing a background inside onboarding itself, essentially baking in a stripped down version of their product to get users to an Aha! moment prior to showing pricing.

...this is the way!

My magic mushroom longevity experiment found a metabolic reset button in the brain. We expected brain changes, but not a potential metabolic breakthrough.

The discovery: my blood glucose control dramatically improved.

+ before: top 2% of population
+ after: top 0.2% of population
+ better than 99.75% of 18-25 yr olds
+ mean glucose dropped 8%
+ time above 125 mg/dL: zero min
+ variability dropped 11%

This single session reduced my estimated HbA1c by 6.8% from 4.7% to 4.4%.

Why does this matter? We treat diabetes and metabolic dysfunction with chronic daily medication (Metformin, Insulin, GLP-1s). This data suggests that a neuroplastic event might have downstream effects on the liver and pancreas that mimic or exceed these drugs.

For context, taking Metformin daily takes 6 months to drop it 10-15%.

This is a first in-human observation, with comparable pre-clinical evidence existing only in rodents.

it took me 5 years to make my first $15k/month from apps🥴

If i had to do it again, I could do it in a few months

here's exactly what i would do

1. launch apps users ACTUALLY want not apps YOU want

"but how do i know what users want?"

easy

get inspiration from other successful apps, find gaps, add your own personality, read the reviews and listen to their users

2. after launching don't touch the app again until you reach at least $1k/month

"but how?"

easy

OBSESS over distribution, don't do anything else, don't even talk to your mom if you have to UNTIL you start developing a viral sense

again, take inspiration from your competition, give their formats a try and iterate fast until you've something that works

3. after reaching $1k/month double down on your winning content and scale to AT LEAST $5k/month

if you found your winners don't waste time testing other things, go ALL IN, post 15 different variations of the same video a day if you have to

4. After reaching $5k month scale with either UGC, Influencers or Paid Ads, the trinity of distribution

The best apps are doing all at once but i'd recommend starting with only one

I went the paid ads route but I know of very successful founders doing the other 2 as well

Two conversations this weekend make me think that there's a vibe shift afoot in Silicon Valley around what one should work on and what is worthwhile.

Culturally, it feels like the moment is ripe for new frameworks:

• Davos expert morality is stale and discredited.
• It's also apparent that the "just be super based" Counter-Enlightenment is not really an answer. (Yes, woke went too far, but simply inverting it doesn't work.)
• EA is no longer the automatic default for smart people.
• There's increasing skepticism of slot and slop machine dynamics.

Overall, "what is worthy and valuable?" feels like it's becoming more central.

Ex-CIA behavior expert Chase Hughes dropped the two simplest, most powerful rules ever created to completely neutralize psychopaths, manipulators, and toxic people — and once you hear them, you’ll never interact the same way again.

In this 36-second clip he reveals:
Always stay visibly calmer than the other person — never escalate, no matter what they throw at you.
Become 100% non-reactive — no shock, no facial reaction, just calm acknowledgment.

His exact line: “In any conversation, the person who reacts the least is the one in charge.”

Thousands are already calling this the single most useful psychology clip they’ve seen in years — it instantly changes how you handle difficult coworkers, narcissists, exes, or even family.

Watch the 36-second clip below and save it. You’ll be glad you did.

🚨 BREAKTHROUGH: A new holographic Transformer just crushed classic HRR memory systems.
The graph says it all — HRR collapses after ~125 items, while DePaola-5 keeps perfect recall up to 509.
That’s a 4× jump in usable associative memory inside a Transformer, with lower noise, higher stability, and zero extra parameters.

The DePaola Transformer is a new kind of attention mechanism that swaps out HRR-style binding for a holographic key transformation that barely accumulates noise. Instead of one fragile channel, each key explodes into several coordinated “holographic” features that reinforce the real signal while canceling interference. The result? A memory system that stays clean and stable hundreds of steps longer than classic HRR. In tests, HRR fell apart at ~125 stored associations, while DePaola-5 held strong all the way to 509 with 4× higher high-SNR capacity — same vectors, same dimension, just a smarter transform. It turns a Transformer’s attention stream into something closer to a noise-resistant working memory rather than a pileup of corrupted sums. It’s simple, fast, parameter-free, and slots straight into any architecture. Think of it as giving Transformers a “lossless memory mode” — a small change with shockingly large impact.

Sauna is one of the most effective health protocols I've done. Here is everything I've learned; it's the most robust characterization ever produced.

Results:

1) Fifteen sessions of sauna dramatically reduced environmental toxins in my body:

+ 65% drop in 2,4-D
+ 100% drop in MEP
+ 15% drop in MBP
+ 100% drop in MEHP (undetectable post sauna)
+ 56% drop in NAPR
+ 56% drop in HEMA
+ 100% drop in Perchlorate (undetectable post sauna)

2. Sauna eliminated 85% of microplastics from my ejaculate.

Nov 2024:
165 particles/mL
July 2025: 20 particles/mL

Nearly identical drop in my blood same time period:
Oct 2024: 70 particles/mL
May 2025: 10 particles/mL

3. Sauna, without ice on the boys, devastated my fertility markers.

Total Motile Count: –56%
Concentration: –30%
Motility: –50%
Morphology: –48%
Count: –9%

4. Sauna coincided with my fertility markers being at an all-time high. I have more total and motile sperm than 99.6% of men of any age, including men under 25.

+ total count: 600 M
+ concentration: 162 M
+ motility: 55%
+ total motile count: 330M
+ morphology (normal): 10%

We do not know what to make of these improvements. Was it the sauna? Sauna + ice? Ice only? We don't know but we did not identify any other protocols or lifestyle changes during this interval that would plausibly account for the change.

5. My vascular function improved by a ten year reduction in age. Now I have the vascular age of an elite 18-early 20s.

+ Central Systolic Blood Pressure: 96 mmHg
+ Central Pulse Pressure: 20 mmHg
+ Pulse Pressure Amplification: 160%
+ SEVR: 227%
+ Augmentation Pressure: 1 mmHg
+ Augmentation Index Wave: 3%
+ Traditional blood pressure: 107/75 mmHg

6. What type of sauna?
Use a dry sauna with high temperatures between 80-100°C (176 to 212°F) and 5-20% relative air humidity. Aim for the lower end of this spectrum, especially as a beginner. Higher temperatures closer to the boiling point can cause side effects like headaches and severely dried nose and eyes.

Note: Steam baths, hot tubs, and infrared saunas fail to replicate the same effects because they do not allow you to safely reach the required high temperatures and do not induce the same level of sweating, the necessary inverted (skin-to-core) temperature gradient, and the massive re-direction of blood to the skin with resulting vasodilation.

Dry sauna is unique, and very likely superior to wet (steam bath) and infrared saunas. By heating up your skin way faster than your core, dry hot sauna flips your core skin temperature gradient, eliciting the following hormetic benefits:

+ enhanced blood flow: the heart pumps up to 70% more blood, similar to intense aerobic exercise (zone 2-+ increased sweating for detoxification: to maintain a stable core temperature, the skin produces 0.6-1 liter of sweat per hour, facilitating significant detoxification.
+ improved heat tolerance: the body becomes better at handling heat, leading to a lower core body temperature (offering metabolic advantages)
+ safe activation of heat shock proteins: the skin experiences substantial heat shock protein activation, while a modest 1°C increase in core temperature is sufficient to activate these proteins without the risk of hyperthermia.
+ extended Exposure at higher temperatures: dry saunas are more tolerable for longer durations and at higher temperatures, maximizing the benefits.

7. Sauna protocol and frequency
Type: hot dry sauna
Temperature: 176–212°F ( I do 200°F)
Relative air humidity: very low, 5-20%
Duration: 20 min
Frequency: 4–7x a week

8. Heat Protection

If you'd like, you can protect your head from the heat by wearing a sauna hat or wrapping it with a towel (use only cotton or other 100% natural material). You can breathe through a towel or cloth if needed to protect your nose. I am personally fine not doing this.

Most importantly, ice the balls.

Icing the testicles is absolutely required to prevent heat from damaging fertility markers.

+ Ice the testes during the sauna session.
+ Use a non-toxic, reusable ice pack material.
+ Wear cotton boxers and shorts.
+ Place ice packs in between the boxers and shorts.
+ Keep them in place for the entire session.

Men should care about preserving fertility markers even when they are not trying to conceive. Sperm quality is tightly coupled to testicular function, which governs testosterone production, metabolic health, and long term endocrine stability. When fertility parameters decline, the same upstream dysfunction often drives lower testosterone, higher inflammation, and increased cardiometabolic risk.

9. Hydration

Dry sauna induces sweating as part of its beneficial mechanism. Be sure to hydrate properly. In general, you might need to rehydrate with up to 16–32 oz (0.5–1 L) of fluid after a sauna session. Be sure to add electrolytes.

If you want to be precise, measure your sweat amount and electrolytes (saltiness) using a patch (e.g., from Gatorade) to quantify your liquid and electrolyte loss, and rehydrate accordingly. Some people have saltier sweat than others and must ensure they replenish electrolytes as well as water.

My results: my body sweats 18 oz during a 20 min sauna at 200 °F, with a sodium concentration of 25-39 mg/oz. A single sauna session flushes 450–700 mg of sodium out of my body.

#

I wish you all the best in life my friend. A new era or being human is here. One where existence is the highest virtue. Prioritize sleep, daily exercise and eat well and you'll be in a strong position. Try to avoid the bad stuff. Anything that takes away your agency.

Be a warrior and caretaker of existence. Don't Die.

The new Holographic Transformer is a rethink of attention itself: instead of forcing every token to compare with every other token, it builds a holographic memory where information is stored and retrieved through smooth mathematical features—cosines, sines, curvature, and phase harmonics. Each key is transformed into a multi-channel signature that behaves like a tiny “wave pattern,” and the model learns by letting these waves interfere, reinforce, or cancel out. This creates a continuous memory flow, not a discrete lookup. The result is a transformer that keeps long-range structure with far less noise, preserves style and context more faithfully, and scales attention linearly instead of quadratically. It’s like giving the model a stable internal geometry for thought—where meanings resonate, not collide. In practice it’s faster, smoother, more coherent, and closer to a genuine holographic working memory than anything built from standard dot-product attention.

In this article, for live-streaming my mushroom trip, I'm harpooned by several psychedelic insiders for "ego enrichment", "a circus of self-indulgence" and "digital narcissism".

Here's what's going on: when someone violates a tribe's norms, the tribe's priests don't try to understand the new behavior, they seek to delegitimize the actor in order to protect their own status and try and claim superiority.

By doing this they are confessing new boundaries of possibilities have been set and their inability to take a similar risk. If they can't win on position or strength, they try to win on virtue.

When you're out on the frontier, expect this kind of behavior. Humans do it without realizing they're doing it. Humans are NPCs. We all are. We're lucky if we can escape NPC mode for a fraction of our life.

Go build something cool. Redefine boundaries. Push past people's comfort zone. When people squawk, understand they're NPCs chirping but it's really just a song they're singing called progress.

“To win this war, everybody has to sit down and think: What can I do?”

In October, we hosted Europe’s biggest Defense Tech Hackathon together with TUM Venture Labs for the 4th time in Munich 💥

Over 300 participants worked across 56 projects, with a special focus on counter-drone solutions to protect our airspace, from detection to classification and interception.

A huge thank you to our key partner @HelsingAI, our hackathon partner Lateration, and everyone else, including TYTAN Technologies, Pilotix, IABG, Autonomous Teaming, Bifrost Defence, SE3 Labs, and the Ukrainian Ministry of Defense @DefenceU for supporting us and making this possible!

Special thanks go to TUM Venture Labs for hosting us — especially Connor Rehn, Florian Doetzer, Antoine Leboyer — and to German Stelmakh 🇺🇦 & Artem Melnychuk (DroneAid Collective), Jonatan H. Luther-Bergquist @mountbranch (EDTH/@inflection_fund), Paolo Trecate & @AkimKarpach9 (EDTH) for co-organizing.

Excited to be back in Munich in February for one of our most important hackathons next year, in parallel to the Munich Security Conference @MunSecConf!

Video credit: @AkimKarpach9

Three exciting longevity results came back from my first dose of mushrooms: >35% decrease in inflammation, a deeply relaxed state and neuroprotective effects.

Inflammation (hsCRP)
My systematic inflammation levels dropped from elite to undetectable. This decrease is significant, aligning with clinical reports of reduced inflammation markers (CRP, TNF-alpha, and IL6) seven days post high dose psilocybin.

+ pre-mushrooms: 0.23mg/dL
+ post-mushrooms: < detection limit 0.15mg/dL
+ >35% decrease

Deep low stress state (HPA Axis)
My cortisol and DHEA-S dropped by 42% and 45%, respectively, suggesting a profound inhibition of the Hypothalamus-Pituitary-Adrenal (HPA) axis, which corresponds with the "after-glow" effect.

+ DHEA-S: 287µg/dL to 157 µg/dL
+ morning cortisol: 10.3 µg/dL to 5.9 µg/dL

While cortisol spikes acutely during the trip, the post-session drop indicates a shift from sympathetic to parasympathetic dominance. This aligns with my sustained relaxed and joyful state.

Neuroprotection (Estradiol)
Estradiol increased 3x (11.3 to 36 pg/ml). This may be due to peripheral activation of 5HT2A receptors stimulating aromatase activity, the enzyme that converts testosterone to estradiol.

A study has shown 5HT2A stimulation in human placental cells can drive aromatase gene expression and activity. My testosterone levels remained unchanged, indicating my body compensated for the minor conversion to estradiol.

The increase is within the normal male range and is generally positive, as estradiol provides anti-inflammatory, neuroprotective and cardiovascular protective effects.

Conclusion
Taken together, these data suggest that a single high dose psilocybin experience can transiently push my biology into a low inflammation, low stress configuration that is theoretically favorable for longevity.

Measurement window: 4 days before vs. 5 days post-dose

NVIDIA CEO Jensen Huang said @elonmusk was there for him as a customer when nobody else was.

“When I announced this thing, nobody in the world wanted it. I had no purchase orders—not one. Nobody wanted to buy it. Nobody wanted to be part of it, except for Elon. He was at the event, and we were doing a fireside chat about the future of self-driving cars. And he goes, “You know what? I have a company that could really use this.” And I said, “Wow, my first customer.” I was pretty excited about it. And he goes, “Yeah, we have this company. It’s a non-profit company”, and all the blood drained out of my face. I had just spent a few billion dollars building this thing. It cost $300,000. And the chances of a non-profit being able to pay for this thing were approximately zero, and he goes, “You know, it’s an AI company, and it’s a non-profit. And we could really use one of these supercomputers.”

So I picked it up. I built the first one for ourselves—we’re using it inside the company. I boxed one up, drove it up to San Francisco, and delivered it to Elon in 2016. I walked up to the second floor, where they were all kind of crammed into a room smaller than this place here, and that place turned out to be OpenAI.”

My biggest learnings from Jeanne DeWitt Grosser (ex-Chief Business Officer at @Stripe, now @Vercel COO):

1. What failed seven years ago now works with AI. In 2017, Jeanne tried to build a system at Stripe that would automatically personalize outbound emails based on company data. Despite working with world-class data scientists, it failed due to too many errors. Today, that exact same approach works. This shows how AI has made previously impossible ideas suddenly viable.

2. A single GTM engineer at Vercel reduced a 10-person sales team to 1 (in just 6 weeks). Jeanne’s team at Vercel had an engineer build an AI agent that handles inbound lead qualification, outbound prospecting, and deal loss evaluation. The agent costs $1,000 per year to run versus over $1 million in salaries for the sales team. The nine displaced team members moved to higher-value work rather than being laid off, and the remaining salesperson is 10 times more efficient.

3. Their AI deal-loss bot has become better at understanding what went wrong than humans. When Jeanne analyzed her biggest loss of the quarter, the salesperson blamed pricing. But an AI agent reviewed every email, call transcript, and Slack message and discovered the real reason: they never spoke to the person who controls the budget, and when ROI came up, the customer clearly didn’t believe the value claims. They are now using AI to analyze sales calls in real time and send alerts like “You’re halfway through the sales process and haven’t talked to a budget decision-maker yet.”

4. Wait until $1 million in revenue before hiring your first salesperson. Founders should continue selling themselves until they reach around $1 million in annual revenue with a repeatable process. The key is having a defined ideal customer profile—customers who look alike.

5. Segment customers on what drives their buying decisions, not just company size. OpenAI has roughly 3,000 employees, which would typically put them in the “mid-market” category. But they’re a top-25 website globally by traffic, so Vercel treats them as enterprise customers requiring complex sales. Effective segmentation combines company size with growth rate, web traffic, workload type, and industry—because selling to e-commerce companies requires completely different language than selling to crypto companies.

6. Most customers buy to avoid risk, not to gain opportunity. About 80% of customers purchase to reduce pain or avoid problems, while only 20% buy to increase upside. This means you should focus your sales messaging on what could go wrong without your product—like falling behind competitors or damaging their reputation—rather than just talking about exciting features. This is especially true when selling to larger companies, where individual careers are on the line.

7. Sales teams should be indistinguishable from product managers—for a bit. Jeanne hires salespeople who have such deep product knowledge that if you put one in front of a group of engineers, it should take 10 minutes to realize they’re not a product manager. This credibility allows sales teams to serve as an extension of research and development—a 20-person sales team talks to hundreds of customers weekly and can translate those conversations into product insights at scale.

8. Building your own AI sales tools may beat buying off-the-shelf software. Because AI is so new and every company’s sales process is unique, Jeanne finds that building custom internal agents often delivers more value than buying vendor solutions. A single go-to-market engineer built their deal analysis bot in just two days, perfectly tailored to their specific workflow. These engineers shadow top salespeople to understand their workflows, then build automation that would have taken months or been impossible just a few years ago.

9. Make every sales interaction great, whether customers buy or not. Jeanne replaced boring discovery calls at Stripe with collaborative whiteboarding sessions where customers drew their payment architecture. Many customers had never visualized their own systems before. They left with a useful asset and a feeling of collaboration, regardless of whether they bought. Many returned years later to purchase. Think about your go-to-market process like a product, not just a sales function.

10. Product-led growth has a ceiling—no $100 billion company runs on it alone. While product-led growth (where users can sign up and start using a product without talking to sales) works well for early growth, customers generally won’t spend a million dollars through a self-service flow. Every major technology company eventually builds a sales team for larger deals. The mistake is waiting too long, since building a predictable sales process takes time.