Jianmin Chen
13 followers · 8 following · 2256 views
on the atlas · 237
- Developing creative identity18 savers
- The Feynman Lectures on Physics Vol. I Ch. 9: Newton’s Laws of Dynamics3 savers
- PCBs without the magic smoke, using agents8 savers
- The Sane Asylum - Colossus21 savers
- From Life in Prison to the Eras Tour | The New Yorker1 savers
- The decoupling capacitor…is it really necessary? - Precision Hub - Archives - TI E2E support forums1 savers
- anti akrasia checklist and techniques | meditationstuff3 savers
- BGA fanout | Sierra Circuits1 savers
- 10 Best High-Speed PCB Routing Practices | Sierra Circuits1 savers
- Microsoft PowerPoint - 2004HS_5Layout1 savers
- Making 1,750 programmable PCB badges for Canada's biggest hackathon · Nick Ficara1 savers
- How to Use AC Coupling Capacitors in High-Speed PCBs1 savers
- Hardware Debugging for Reverse Engineers Part 2: JTAG, SSDs and Firmware Extraction | Wrongbaud's Blog1 savers
- Québec - Use & Modify1 savers
- (...) — USB, Thunderbolt, Displayport & docks1 savers
- What are Differential Pairs and Why Do We Use Them!1 savers
- flashrom README — flashrom v2.0.0-devel (git:v1.8.0-82-g092b3fc8) documentation1 savers
- communication - USART, UART, RS232, USB, SPI, I2C, TTL, etc. what are all of these and how do they relate to each other? - Electrical Engineering Stack Exchange2 savers
- ASM2464PDX_DataSheet_unlocked--1--0b2fdebe-8b0c-4c1a-b0a4-694c9e5e019d.pdf1 savers
- Explainer: Thunderbolt – The Eclectic Light Company1 savers
- Finally Taming Thunderbolt With Third-Party Chips | Hackaday1 savers
- PCIe, part 4: the physical layer (1) — Shengyi Wei1 savers
- PCIe, part 3: the data link layer — Shengyi Wei1 savers
- PCIe, part 2: the transaction layer — Shengyi Wei1 savers
- Thunderbolt (interface)1 savers
- Build your own React7 savers
- PCB Manufacturing & Assembly Capabilities - JLCPCB1 savers
- HyperGraph Autorouting - by Seve Ibarluzea - autorouting1 savers
- Designing a Tree Diff Algorithm Using Dynamic Programming and A* - Tristan Hume4 savers
- Katrina Ellison Geltman1 savers
- A* Demystified - Gabriel Gambetta1 savers
- Introduction to the A* Algorithm1 savers
- 13 things I would have told myself before building an autorouter2 savers
- Lecture 2 - What's a PCB?5 savers
- how to make time last longer - Chandhana6 savers
- SGD Optimization Algorithms2 savers
- The Unreasonable Effectiveness of Recurrent Neural Networks19 savers
- So you want to AI your PCBs? — eddie1 savers
- Visualizing and Understanding Convolutional Networks1 savers
- You are probably underestimating how good self-love can be - LessWrong8 savers
- Neural networks and deep learning1 savers
- [1603.07285] A guide to convolution arithmetic for deep learning2 savers
- Why Momentum Really Works9 savers
- Particle は: Topic Marker1 savers
- The Japanese Grammar Index1 savers
- How to Unclench | Jonny Miller23 savers
- Seeing Theory - Basic Probability2 savers
- ET Book · Edward Tufte on GitHub1 savers
- Neural networks and deep learning1 savers
- Nintil - Massive input and/or spaced repetition6 savers
- CS231n Deep Learning for Computer Vision1 savers
- Understanding LSTM Networks -- colah's blog10 savers
- Differentiable Programming from Scratch – Max Slater – Computer Graphics, Programming, and Math5 savers
- PyTorch internals : ezyang’s blog7 savers
- transformer_attention.pdf23 savers
- Making Deep Learning Go Faster32 savers
- Building Machine Learning Systems for a Trillion Trillion Floating Point Operations - YouTube6 savers
- TheEraOfExperience.pdf2 savers
- Jane Street Blog - Can you design a chip? Announcing the protocol emulator ASIC competition1 savers
- A Practical Look at PIO on the Raspberry Pi Pico | Blues Wireless1 savers
- An Overview of Deep Learning for Curious People | Lil'Log4 savers
- Calculus on Computational Graphs: Backpropagation -- colah's blog6 savers
- Intro to Proofs for the Morbidly Curious1 savers
- How To Understand Things - Nabeel S. Qureshi18 savers
- How To Learn Trigonometry Intuitively – BetterExplained3 savers
- Linear Algebra and Multivariable Calculus2 savers
- Matrix_Calc_for_DL.pdf1 savers
- ClawPod: OpenClaw on HomePod | Alexis Gallagher1 savers
- Autodidax: JAX core from scratch — JAX documentation2 savers
- A visual proof that neural nets can compute any function5 savers
- Rest in motion24 savers
- Quickstart: How to think in JAX — JAX documentation2 savers
- BiereCimattiClarkeStrichmanZhu-Advances-58-2003-preprint.pdf1 savers
- On Self-Respect – Alex Burgin16 savers
- Engineer 'builds a GPU from scratch' in two weeks — process much harder than he expected | Tom's Hardware1 savers
- Z3 Internals (Draft)1 savers
- Abdul Rahman Sibahi | A Dumb Introduction to z31 savers
- SAT Solvers: An Introduction and Go Implementation1 savers
- Command Reference - YosysHQ Yosys 0.68-dev documentation1 savers
- NP-Completeness1 savers
- KLayout Layout Viewer And Editor1 savers
- fpgas.online — Access Real FPGAs From Your Browser1 savers
- rachel2020regds.pdf1 savers
- Application Specific Integrated Circuit (ASIC) Floorplan Automation - Part I | MakinaRocks Tech Blog1 savers
- The high-return activity of raising others' aspirations - Marginal REVOLUTION3 savers
- Spaced Repetition for Efficient Learning · Gwern.net4 savers
- nspfssd-lisa2004-final.pdf1 savers
- Using Nix with Dockerfiles – Mitchell Hashimoto1 savers
- LNCS 6174 - ABC: An Academic Industrial-Strength Verification Tool1 savers
- "Modular Electronics Learning (ModEL) project"1 savers
- Layman's Guide to IC Reverse Engineering3 savers
- A primer on digital circuit synthesis - YosysHQ Yosys 0.68 documentation1 savers
- 2025 letter | Zhengdong33 savers
- Reverse-engineering the 8086's Arithmetic/Logic Unit from die photos1 savers
- WaveDrom - Digital timing diagram everywhere1 savers
- Inside the vintage 74181 ALU chip: how it works and why it's so strange1 savers
- Hacker Fab Documentation | Hacker Fab2 savers
- PDK | Zero to ASIC Course1 savers
- startup review: axiom math - erika milktea1 savers
- Inside the 74181 ALU chip: die photos and reverse engineering1 savers
highlights · 326
We can compensate for this by sampling several angles and averaging the results8.
GPU text rendering with vector textures · Will DobbieThere are kanji that only have one or the other.
Onyomi vs. Kunyomi: What's the Difference?Do not communicate by sharing memory; instead, share memory by communicating.
Effective Go - The Go Programming LanguageAs the disk spins, the write head very precisely controls its magnetic field to alter the polarity of the magnetic grains, essentially flipping them.
Making Software: How is data stored?Well, it's covered in a very thin magnetic layer that is divided into hundreds of thousands of concentric circles around the surface called tracks. Each track is divided into sectors that have a specific structure, with the first chunk of data being used as a preamble to make sure that the read/write head knows how fast the disk is spinning and therefore how often to sample the magnetic signal.
Making Software: How is data stored?To mitigate this the memory controller will look for gaps to opportunistically refresh a bank when the CPU isn't currently requesting anything from it.
Making Software: How is data stored?Dynamic random access memory
Making Software: How is data stored?To mitigate this, we can split the caches that use SRAM into smaller banks - a 1024kb cache can be split into 8 smaller 128kb banks to use less power.
Making Software: How is data stored?Out of these switches we can construct logic gates13 and the particular logic gate we need to create some memory is an inverter or a NOT gate, which inverts the signal applied to it. If there is a voltage applied to the inverter, it returns 0v and if there is no voltage it returns a voltage.
Making Software: How is data stored?We just built a logic inverter! It takes the logical value from our input signal and inverts it (high becomes low, low becomes high). If we encode "0" bits as low voltage and "1" bits as high voltage, we just built a one-bit binary NOT gate.
Computing with Transistors - Andrew GibianskyIn a durability review, we encourage the same “what are all the things that might go wrong” thinking, and really encourage engineers to be creatively critical of their own code.
Building and operating a pretty big storage system called S3 | All Things DistributedFor example, we use an algorithm, such as Reed-Solomon, and split our object into a set of k “identity” shards. Then we generate an additional set of m parity shards. As long as k of the (k+m) total shards remain available, we can read the object. This approach lets us reduce capacity overhead while surviving the same number of failures.
Building and operating a pretty big storage system called S3 | All Things DistributedBut as we aggregate millions of workloads a really, really cool thing happens: the aggregate demand smooths and it becomes way more predictable.
Building and operating a pretty big storage system called S3 | All Things DistributedAmazon's distributed storage
NYSRGBeginning Kanji & Stockpiling Kanji Knowledge
Learn Japanese: A Ridiculously Detailed GuideYour goal should be to learn 20-30 kanji and ~100 vocabulary words that use those kanji (and only those kanji) each week.
Learn Japanese: A Ridiculously Detailed Guide"여러 변덕스러운 우연이, 지쳐버린 타인이, 그리고 누구보다 자신이 자신에게 모질게 굴 수 있으니 마음 단단히 먹기 바랍니다. 나는 커서 어떻게 살까, 오래된 질문을 오늘부터의 매일이 대답해줍니다. 취업 준비, 결혼 준비, 육아 교육 승진 은퇴 노후 준비를 거쳐 어디 병원 그럴듯한 일인실에서 사망하기 위한 준비에 산만해지지 않기를 바랍니다. 무례와 혐오와 경쟁과 분열과 비교와 나태와 허무의 달콤함에 길들지 말길, 의미와 무의미의 온갖 폭력을 이겨내고 하루하루를 온전히 경험하길, 그 끝에서 오래 기다리고 있는 낯선 나를 아무 아쉬움 없이 맞이하길 바랍니다." - 허준이 교수님
For the Years to ComeThis came with its own ups and downs as balancing curiosity and exploration with getting things done has been a challenge of its own. But I've been asking myself if I can do more than just to continue exploring beautiful puzzle pieces in my own world.
For the Years to Comeundergo the artist’s rite of passage — to find "a really shitty apartment and move out here and figure it out."
ADÉLA: "My whole ethos is that I'm extremely imperfect...and I find that so beautiful" | The Line of Best FitI want to be 2006 Lady Gaga in the parking lot doing 'Just Dance'. That’s my vibe I want to be on.
ADÉLA: "My whole ethos is that I'm extremely imperfect...and I find that so beautiful" | The Line of Best Fit"That would have been nice to hear," she says.
ADÉLA: "My whole ethos is that I'm extremely imperfect...and I find that so beautiful" | The Line of Best FitNo matter what you do in that situation, [the audience is] just gonna keep on spinning their own narrative about it.
ADÉLA: "My whole ethos is that I'm extremely imperfect...and I find that so beautiful" | The Line of Best Fit"The only music that I've ever listened to is pop music. Like, what am I doing?"
ADÉLA: "My whole ethos is that I'm extremely imperfect...and I find that so beautiful" | The Line of Best Fitand is not shy about the fact that it hit her hard: "It was the worst year of my life," she says, calling it "identity-stripping."
ADÉLA: "My whole ethos is that I'm extremely imperfect...and I find that so beautiful" | The Line of Best FitI didn't want to say it out loud, because I knew it was possible for me. I didn't want [my parents] lack of belief to affect my belief in myself, and so they just were not aware. And then one day I was speaking English fluently when I was nine. They were, like, 'Okay, I guess you have a special interest.'
ADÉLA: "My whole ethos is that I'm extremely imperfect...and I find that so beautiful" | The Line of Best FitMany individual requests and streams are multiplexed over a single TCP connection. Kubernetes sets a default HTTP/2 limit of 100 concurrent requests (or technically streams) per TCP connection.
k8s-1m OverviewThin provisioning is a way for a storage device to pretend to have more space than it actually does, relying on the administrator to add more storage "just in time".
Thin provisioning [LWN.net]Well, pretty much all of conventional lens optics — eye glasses, contact lenses, microscopes, cameras, telescopes, magnifying glasses — is based on refraction.
Optics basics: refraction | Skulls in the StarsIn other words, the speed of light in vacuum is meters/second, and the speed of light in the medium is .
Optics basics: refraction | Skulls in the StarsThese concerns can be viewed as two separate layers of indirection stacked on top of one another: (1) the guest creates page tables to describe how to map guest virtual addresses (gVAs) to guest physical addresses (gPAs), and (2) the VMM creates mappings from gPAs to host physical addresses (hPAs).
The evolution of an x86 virtual machine monitorThough jitting C is not feasible, one can find a middle ground through Profile Guided Optimization (PGO, cutely [and uncommonly] pronounced “pogo”).
A Deep Introduction to JIT Compilers: JITs are not very Just-in-time | kipply's blogMost JITs however (Pypy, Java, JS Engines), are not at all about compiling code just-in-time, but compiling optimal code at an optimal time.
A Deep Introduction to JIT Compilers: JITs are not very Just-in-time | kipply's blogTo improve performance, the x86_64 architecture caches the last few translations in the so-called translation lookaside buffer (TLB).
Introduction to Paging | Writing an OS in RustJITs improve implementations in speed by being able to optimise (compile) on information that is only available at runtime.
A Deep Introduction to JIT Compilers: JITs are not very Just-in-time | kipply's blogTo speed up inter-CCF transfers, our translator, like pre- vious ones [9], employs a “chaining” optimization, allowing one CCF to jump directly to another without calling out of the translation cache (TC).
The evolution of an x86 virtual machine monitorThe translator provides the ability to intercept precisely the virtualization sensitive instructions without re- quiring trap semantics. Since we cannot patch instructions in the guest’s own image (such patching would be visible to the guest), we must translate all guest kernel code and always execute from a guest-invisible translation cache.
The evolution of an x86 virtual machine monitorSince cli is a privileged instruction, it must be handled spe- cially by the VMM. Unlike popf , cli can generate a trap, but it is more efficient to avoid the trap by translating it non-identically.
The evolution of an x86 virtual machine monitorThe translator accumulates IR objects into a translation unit (TU), stopping at 12 instructions or a terminating in- struction (usually control flow).
The evolution of an x86 virtual machine monitorFor example, VMware’s vSphere ESX hypervisor is com- prised of the vmkernel and a VMM.
The evolution of an x86 virtual machine monitorThe attempt to address this is virtio: a series of efficient, well-maintained Linux drivers which can be adapted for vari- ous different hypervisor implementations using a shim layer. This includes a simple extensible feature mechanism for each driver.
virtio-paper.pdfthe agency discussion often labels people as agentic vs non-agentic, which i find very unhelpful because how many people are blocked by psychological issues like self-sabotage and internalized pessimism and hypercompetitive frameworks?
eating the fire - by vincent huang - a slice of my mindi guess because of some superstition about jinxing / the world coming after people who declare themselves too boldly, which is just stupid? literally what is the point. i have eventually solved every personal problem i’ve ever encountered and there’s no reason to keep pretending otherwise
eating the fire - by vincent huang - a slice of my mindi’m a case of extreme resource mismanagement. not to say that my life is bad right now, but i finally understood that my life could be so much better if i actually managed my assets properly.
eating the fire - by vincent huang - a slice of my mindThis is one drawback to CRDTs — we can only ever add information, not remove it. Although from the application’s perspective the key has been fully deleted, the underlying state still records that the key was once there. In technical terms, we say that CRDTs are monotonically increasing data structures.
An Interactive Intro to CRDTs | jakelazaroff.comA register is a CRDT that holds a single value. There are a couple kinds of registers, but the simplest is the Last Write Wins Register (or LWW Register).
An Interactive Intro to CRDTs | jakelazaroff.comState-based CRDTs transmit their full state between peers, and a new state is obtained by merging all the states together. Operation-based CRDTs transmit only the actions that users take, which can be used to calculate a new state.
An Interactive Intro to CRDTs | jakelazaroff.comKahneman continued to collaborate with Tversky. But he also took pains to distance himself from this man, with whom he’d once shared a typewriter in a small office in Jerusalem. The ill feelings wouldn’t ease up until Tversky told Kahneman he was dying of cancer in 1996.
It’s Your Friends Who Break Your Heart - The AtlanticAre you unmoved to write to me because your meds have worked so well that you’re now perfectly functional, to the extent that you need not go searching for ways to narrate/make sense of your internal landscape?
It’s Your Friends Who Break Your Heart - The AtlanticIt’s a painfully familiar dynamic in a friendship: One friend says, Get a grip already. And the other one says, I’m trying. Can’t you see I’m trying? Neither party relishes her role.
It’s Your Friends Who Break Your Heart - The Atlantic“We were sort of in that thing where you’re like, ‘You’re my savior,’ ” Rebecca told me. “Like, you cling to each other, because you’ve found each other.”
It’s Your Friends Who Break Your Heart - The Atlantic