Trevor Trinh
19 followers · 33 following · 15655 views
on the atlas — 233
- Multicast1 savers
- The 7 Minute Cleaning Miracle. The Shinkansen cleaning service… | by IGNITION Staff | IGNITION INT. | Medium1 savers
- Questions · Patrick Collison26 savers
- What Should You Do with Your Life? Directions and Advice - Alexey Guzey74 savers
- Every productivity thought I've ever had, as concisely as possible - Alexey Guzey30 savers
- How to be more ambitious: A tutorial for humans with models27 savers
- Demanding and Supportive | RKG2 savers
- The Confrontation Obligation | RKG1 savers
- An idiot in motion goes farther than a genius at rest | Alexandria6 savers
- “Ask for what you need” is a terrible social system :: the6thspence.substack.com/14 savers
- Redesigning source control at Anthropic \ Anthropic10 savers
- How to Build Wealth as a Career Person – Asher's Blog2 savers
- Terms & Conditions | Blink1 savers
- MEV | Blink1 savers
- Code was our medium for thought8 savers
- Options 101: Free options education – Options 1012 savers
- Kokoro TTS: Free Text-to-Speech in Your Browser - Kokoro Web1 savers
- Rare UI — Rare Animated React Components2 savers
- A Deep Dive into CLOB Microstructure - by Annanay Kapila1 savers
- Your First HFT Alpha: From Flow Prediction to Order-Book Imbalance1 savers
- From Linear Alphas to Conditional Alpha: Decision Trees, Random Forests, and the Models Behind Modern HFT1 savers
- Advice to Young People, The Lies I Tell Myself - jxnl.co19 savers
- GLM-5.3 and the spread of advanced cyber capabilities \ Anthropic7 savers
- Herdr — one terminal for the whole herd3 savers
- How Notion handles concurrent editing with CRDTs7 savers
- Nat Friedman60 savers
- Figure Technology: Smoke And Mirrors From A Fast-And-Loose Lender Masquerading As A Blockchain Darling1 savers
- Previewing the Model Hardware Standard \ Anthropic10 savers
- Dockview1 savers
- Same Same But Different - by Arthur Hayes1 savers
- Thomas Peterffy2 savers
- The Making of a Market Maker - Colossus3 savers
- Sarah Guo's Wager - Colossus16 savers
- The New World - Colossus7 savers
- RLHF Book by Nathan Lambert15 savers
- Hedge What You Can't Control — Castle1 savers
- On the Future of Hedging — Castle1 savers
- How a California goat herding company is hedging against risk of higher wages1 savers
- Musings on Trading — Castle2 savers
- Getting Started | Helm1 savers
- Bitcoin Battlefield: Live Buy/Sell Wall & Liquidation Map | Newhedge1 savers
- If you teach a child to just power through unpleasant tasks, she'll lack self-discipline4 savers
- In Memory of My Wife, Elise Cawley (1961–2026), with Thanks for 36 Wonderful Years—Stephen Wolfram Writings26 savers
- The roll up opportunity of the next five years1 savers
- The Dark Night of Mathematics - by Kirwin Hampshire1 savers
- Learning to Replicate Expert Judgment in Financial Tasks - Thinking Machines Lab5 savers
- Making 768 servers look like 1 — PlanetScale9 savers
- Re: Linking Patchwork with Sashiko? - Linus Torvalds1 savers
- What Are Namespaces and cgroups, and How Do They Work? – NGINX Community Blog1 savers
- Rewriting Bun in Rust | Bun Blog12 savers
- Expertise Acceleration - Commoncog1 savers
- Where Are the Good AI Products? - by Varun Shenoy1 savers
- Chinese_numerology1 savers
- Fast and Simple Rust Interner1 savers
- Markouts: Measuring Adverse Selection | Udit Samani1 savers
- Math Academy2 savers
- Profit-Seeking, Self-Interest, And Virtue | Finance and Philosophy1 savers
- Entrepreneurship: Productive, unproductive, and destructive - ScienceDirect1 savers
- The Agent Skills Directory2 savers
- The Mathematical Reason Most People Never "Make It"2 savers
- Mini Blog Post 24: On Procrastination - The art of shaping your future actions — Neel Nanda3 savers
- Evan Chen • Napkin (v1.5)10 savers
- Post 34: Learning how to learn — Neel Nanda20 savers
- Mini Blog Post 19: On systems - Living a life of zero willpower — Neel Nanda16 savers
- Mini Blog Post 23: Taking Social Initiative — Neel Nanda10 savers
- Mini Blog Post 3: Become a person who Actually Does Things — Neel Nanda45 savers
- Post 35: Your Standards are Too High — Neel Nanda7 savers
- The Intellectual Obesity Crisis - by Gurwinder - The Prism4 savers
- How to be More Agentic - by Cate Hall - Useful Fictions24 savers
- Mini Blog Post 11: Live a life you feel excited about — Neel Nanda10 savers
- Post 33: Seek Upside Risk — Neel Nanda6 savers
- Post 42: When You Already Know the Answer — Neel Nanda5 savers
- the art of reading tea leaves4 savers
- School Is Not Enough25 savers
- Make Something Heavy - by Anu - Working Theorys17 savers
- What Aristotle can teach us about AI-enabled quantitative investment1 savers
- Jane Street Blog - Formal methods and the future of programming3 savers
- Home | Modos2 savers
- commonware > Constantinople1 savers
- Vector embeddings | OpenAI API1 savers
- Silicon Spin Qubits: How Transistor Tech Becomes Quantum1 savers
- Depository Trust & Clearing Corporation2 savers
- IO devices and latency — PlanetScale1 savers
- C++11 Memory Model1 savers
- Ideas — Noah Zender9 savers
- Introduction | Modal Docs1 savers
- Slurm Workload Manager3 savers
- CBB on X: "How CBB Cartel & Friends Made +$40m on $XPL " / X1 savers
- Tokenized Securities Go Mainstream, Hyperliquid Enters Prediction Markets, and Solana Gathers in Miami | Galaxy1 savers
- Claude Mythos Preview \ red.anthropic.com13 savers
- Arrakis on X: "Who is actually trading on Trade.xyz?" / X1 savers
- advice16 savers
- The end of my childhood17 savers
- About Me2 savers
- Saneel on X: "The Hedonist's Stone" / X3 savers
- marleyx.com/favorite-words2 savers
- Melatonin · Gwern.net2 savers
- How To Be Successful96 savers
- the friendship theory of everything69 savers
- Speed matters: Why working quickly is more important than it seems « the jsomers.net blog68 savers
highlights — 2190
Babel was being constructed before our eyes and there was some mechanism for separating the masterpieces from the random strings of text, and all the masterpieces were dropping as fast as publishers could scoop them up? Would authors stop writing then
The Dark Night of Mathematics - by Kirwin Hampshirehe creation (or even the pursuit) of novel mathematics is one way that humans have historically accessed the ineffable and encountered the divine and mystical
The Dark Night of Mathematics - by Kirwin HampshireIn contrast, we’ve shown that high-quality proprietary datasets labeled by expert investors and used for fine-tuning produce custom models that exceed frontier performance on our tasks
Learning to Replicate Expert Judgment in Financial Tasks - Thinking Machines Labif a model couldn’t match an example from its own training set then either the example is genuinely difficult, or the original label was wrong.
Learning to Replicate Expert Judgment in Financial Tasks - Thinking Machines LabAfter examining the reasoning traces of the model we realized that the labels in the dataset were often wrong. Since expert labelers are costly, we devised a verification scheme that routes only the contested examples to experts
Learning to Replicate Expert Judgment in Financial Tasks - Thinking Machines LabFine-tuning sidesteps this: rather than contorting the expert’s intuition into a static prompt, the training process lets the model develop its own judgment
Learning to Replicate Expert Judgment in Financial Tasks - Thinking Machines LabGPT 5.4 costs 43% more than 5.2 but is only marginally more accurate.
Learning to Replicate Expert Judgment in Financial Tasks - Thinking Machines LabOur proprietary model outperforms all frontier models we tested on information accuracy and recall, at a fraction of their cost.
Learning to Replicate Expert Judgment in Financial Tasks - Thinking Machines LabThis metadata is stored in the router, and tells it that the user table is sharded on its id column using the user_hash shard index. This user_hash shard index uses the router's built-in value hashing. For each incoming row, it hashed the ID, and uses this to send it to the correct shard to be stored.
Making 768 servers look like 1 — PlanetScaleThe router must understand the data topology, create a plan for distributing the query to all shards that may contain matching results, aggregate the results back at the router, and send the full result set to the client.
Making 768 servers look like 1 — PlanetScaleIn the kernel community we do open source because it results in better technology, not because of religious reasons. And so we make decisions primarily based on technical merit. Not fear of new tools.
Re: Linking Patchwork with Sashiko? - Linus TorvaldsReal container support was added to the Linux kernel only in 2013, however. This is what made namespaces really useful and brought them to the masses
What Are Namespaces and cgroups, and How Do They Work? – NGINX Community BlogPre-merge, this took 5.9 billion uncached input tokens, 690 million output tokens, and 72 billion cached input token reads — around $165,000 at API pricing. By hand, I think this would've taken 3 engineers with full context on the codebase about a year, during which time we wouldn't be able to improve Node.js compatibility, fix bugs, fix security issues or implement new features.
Rewriting Bun in Rust | Bun BlogIf you need a paragraph-long comment to justify why the workaround is OK, the code is wrong — fix the code.
Rewriting Bun in Rust | Bun BlogUsually with humans, the person reviewing the code is not the person who authored the code. The person writing the code wants to merge the code, which can bias their actions to ship before it's ready.
Rewriting Bun in Rust | Bun BlogThe main problem with DP is that coming up with DP exercises is not free — someone has to spend time developing effective DP exercises for your skill domain.
Expertise Acceleration - CommoncogIdeas without the right maturation of capabilities won't find product-market fit. The key is recognizing which applications are on the cusp of viability at a given moment based on the evolution of the technology
Where Are the Good AI Products? - by Varun ShenoyThe Chinese interpretation of 4 as unlucky is a more recent development, considering there are many examples, sayings and elements of the number 4 considered as auspicious instead in Chinese history.[2]
Chinese_numerologyInterning works by ensuring that there’s only one canonical copy of each distinct string in memory. It can give the following benefits:
Fast and Simple Rust Internerany counterparty who predictably trades in the direction of future price moves is informationally advantaged. The markout measures this regardless of why.
Markouts: Measuring Adverse Selection | Udit SamaniBackground volatility 𝜎 σ, by contrast, does not shift the expected markout curve at all. Larger 𝜎 σ just makes the distribution wider, requiring more trades to reliably estimate the mean.
Markouts: Measuring Adverse Selection | Udit SamaniThe crossover time 𝜏 ∗ = − 1 𝜆 ln ( 1 − 𝑠 / ( 2 𝛼 𝜇 ) ) τ ∗ =− λ 1 ln(1−s/(2αμ)) is the window you have to hedge before a toxic fill turns into a loss
Markouts: Measuring Adverse Selection | Udit SamaniConcentrated toxicity suggests a specific adversarial counterparty; diffuse toxicity suggests the market structure itself has changed
Markouts: Measuring Adverse Selection | Udit Samaninet edge (gross minus adverse selection). The markout at long horizon is the net edge
Markouts: Measuring Adverse Selection | Udit SamaniIn practice, you fit 𝛼 𝜇 αμ and 𝜆 λ from the empirical curve, then directly use 𝛼 𝜇 αμ to set your minimum spread:
Markouts: Measuring Adverse Selection | Udit SamaniThis crossover time defines your safe holding horizon: if you can fully hedge your inventory before 𝜏 ∗ τ ∗ , you escape the adverse selection
Markouts: Measuring Adverse Selection | Udit SamaniThe spread exceeds the long-run adverse selection cost. You earn positive expected PnL on every trade
Markouts: Measuring Adverse Selection | Udit SamaniPnL starts at 𝑠 / 2 s/2 and decays to a positive but smaller long-run value. You still make money, but less than the gross spread suggests
Markouts: Measuring Adverse Selection | Udit Samanithe same spread is worth less when informed traders move prices more.
Markouts: Measuring Adverse Selection | Udit Samaniquickly the informed signal appears in the mid price. Large 𝜆 λ: information is revealed quickly, the adverse selection hits you fast
Markouts: Measuring Adverse Selection | Udit Samaniinformation cost: the fraction of trades that are informed, multiplied by how much each informed trade ultimately moves the price
Markouts: Measuring Adverse Selection | Udit SamaniThe mid price will drift upward as their information is incorporated into the market
Markouts: Measuring Adverse Selection | Udit Samanihey trade for idiosyncratic reasons — portfolio rebalancing, hedging, liquidity needs. Their trades carry no information about future price direction. After an uninformed buy, the mid price is equally likely to go up or down
Markouts: Measuring Adverse Selection | Udit Samanimarkouts are often reported from the taker’s perspective
Markouts: Measuring Adverse Selection | Udit Samaniif the price rose after you sold, you lost money on the short position you’re now holding. That price rise is adverse selection — the buyer knew more than you
Markouts: Measuring Adverse Selection | Udit SamaniIt is the simplest diagnostic in market making: does the price move toward you after a fill (good) or against you (toxic)?
Markouts: Measuring Adverse Selection | Udit Samanithe virtuous entrepreneur, recognizing that her rewards are disproportional to her merit, and that the farmers’ compensation is not proportional to the value they create (and to the risks and efforts they undertake) would seek a fairer distribution of the overall value created (for instance, by paying farmers more for their product
Profit-Seeking, Self-Interest, And Virtue | Finance and PhilosophyOf course, these profit-seekers should get moral credit by profiting only from entrepreneurial profits; it takes moral fortitude to forgo the opportunity to profit through rent seeking (
Profit-Seeking, Self-Interest, And Virtue | Finance and PhilosophyA value-creating padre, recognizing that resources in the POW camp would be much better allocated if prisoners used a bulletin board to exchange their goods, would promote the use of such board even though this would put a stop to his trading bonanza
Profit-Seeking, Self-Interest, And Virtue | Finance and PhilosophyThe opinion towards the padre in the camp was not positive. It was said that he did not improve the traded products but merely “resold’’ them at a higher price, profiting at the expense of others. Munger and Russell argue that this assessment is mistaken. The padre corrects a mistaken allocation of resources, helping those who don’t smoke get cheese in exchange for their cigarettes. Thus, he did not profit at the expense of others, but earned his profits by finding opportunities that created value for others and reducing transaction costs (
Profit-Seeking, Self-Interest, And Virtue | Finance and Philosophypolicy can influence the allocation of entrepreneurship more effectively than it can influence its supply
Entrepreneurship: Productive, unproductive, and destructive - ScienceDirectsociety's entrepreneurial activities varies much more because of their allocation between productive activities such as innovation and largely unproductive activities such as rent seeking or organized crime
Entrepreneurship: Productive, unproductive, and destructive - ScienceDirectThe paradox is you need both. Explore the noise until you have signal. Then exploit that.
The Mathematical Reason Most People Never "Make It"Patterns are emerging. “Oh, this is what works. Huh.” That’s your √n. The late game is exploitation. You double down on the winners. You cut out the losers. You focus your energy towards the best bet
The Mathematical Reason Most People Never "Make It"You can’t know which 10 posts out of 100 will blow up without posting all 100. You can’t know which skills are your multipliers without trying a bunch of skills. You can’t know which relationships matter without meeting a lot of people.
The Mathematical Reason Most People Never "Make It"Some people will compound their advantages. Most won’t. Mathematics doesn’t give a fuck about fairness or ethics. Compounding systems create exponential distributions, that’s it.
The Mathematical Reason Most People Never "Make It"Price’s Law suggests that if you’re consistently in that square root across multiple domains, luck isn’t the full story. You might actually be cooking at a different level, and the isolation you feel—the sense that few people “get it”—is mathematically predictable
The Mathematical Reason Most People Never "Make It"surprise is a pretty visceral emotion, and is good at hacking into my motivation system
Mini Blog Post 24: On Procrastination - The art of shaping your future actions — Neel NandaA flash of actually caring, that’s enough to break the inertia and get me to spend some willpower on it.
Mini Blog Post 24: On Procrastination - The art of shaping your future actions — Neel NandaYou don’t want to be spending willpower on this - any strategy that involves spending willpower in the longterm is not robust
Mini Blog Post 24: On Procrastination - The art of shaping your future actions — Neel Nanda