Search the archive

FIVE BOOKS THAT HAVE SHAPED MY CAREER

A PM's success hinges on more than technical expertise — it's about building trust, measuring impact, and challenging assumptions.

HOW WE BUILT NGROK'S DATA PLATFORM

Ngrok built its data platform with an engineering team of just one person, using open-source tools and self-hosting on Kubernetes. They had challenges like complex schemas and large data volumes, which were addressed through code generation, schema normalization, and scaling Apache Flink with Protobuf messages.

DATA SYNCHRONIZATION

React Query tackles the challenge of data synchronization by keeping the data displayed on the screen up-to-date with the server state. It uses a "Stale While Revalidate" strategy, where data is considered stale by default and refetched in the background when triggers like changes in query keys, new component mounts, window focus, or device going online occur. This allows for a smooth user experience with instant data delivery while still making sure the cache stays synchronized with the server.

PHRASE MATCHING IN MARGINALIA SEARCH

A search engine added phrase matching by transitioning from approximate word position storage to storing precise positions using compressed lists. This change required implementing a new way to store information about word occurrences within different document fields. The ranking algorithm was upgraded to take advantage of the new phrase matching capabilities, incorporating factors like phrase presence, keyword distance, and early occurrence.

WHEN TO DO WHAT YOU LOVE

If your main goal is to make money, you may have to compromise on your passion, as most people pay you for what they want, not what you want. However, if you are passionate about something that is also in high demand, like a niche market, you can potentially earn a living from it. Building a startup can be a viable option here, but mostly if it arises from your own passions and interests.

THE GRUG BRAINED DEVELOPER

This is a list of various pieces of advice on writing good software. For example, good devs fight against complexity, say "no" to unnecessary features, and embrace simple solutions. They are also patient and refactor thoughtfully.

DON'T BE AFRAID TO RE-INVENT THE WHEEL

Relying heavily on external codebases can create risks like security vulnerabilities, dependence on third-party maintenance, and ongoing integration work. It's fine to sometimes replicate external code from packages in internal codebases because this gives developers better control and security over that code.

all resources here

HUMANIFYJS

HumanifyJS is a tool that uses LLMs and Babel to deobfuscate, unminify, transpile, decompile, and unpack JavaScript code, making it more readable for humans.

FACAD

Facad is a colorful directory listing tool for the command line that features emoji-based file and directory icons, sorted output, support for symlinks and executables, and customizable column width.

EVOLUTION API

EvolutionAPI is an open-source WhatsApp integration API that provides a RESTful API service for controlling WhatsApp functions, allowing users to create multiservice chats, service bots, and other systems utilizing WhatsApp.

DO AI COMPANIES WORK?

AI companies face economic challenges due to the high costs of building and maintaining models. As newer and more advanced LLMs emerge, older ones quickly become obsolete, making it easy for a company to be left behind. Now, companies must either spend enormous amounts of money to stay ahead or risk falling behind and losing market share.

EPIC IS SUING GOOGLE — AGAIN — AND NOW SAMSUNG, TOO

Epic Games is suing Google and Samsung. The lawsuit focuses on Samsung's "Auto Blocker" feature, which makes it difficult for users to install apps from third-party stores. Epic argues that Auto Blocker is designed to protect Google and Samsung's app stores, not users, and that it is blocking the growth of the Epic Games Store.

NO SUCH THING AS EXACTLY-ONCE DELIVERY

The difference between delivery and processing in messaging systems is that delivery refers to the act of successfully transferring a message to a receiver, while processing covers the entire message lifecycle, including delivery, processing by the receiver, and acknowledgement. Sequin is a Postgres streaming library that prioritizes exactly-once processing by making sure that every message is delivered, processed, and acknowledged without duplication. Achieving exactly-once delivery is practically impossible due to the two-phase commit problem - confirming the simultaneous confirmation of message delivery across different locations.

BOP SPOTTER

Bop Spotter is an experimental project that is using a modified Shazam app attached to a pole to monitor music trends in San Francisco's Mission district.

ARE YOU SAYING THAT MY CODE IS BORING? THANK YOU!

“Boring” code is the best kind of code because it allows teams to focus on solving business problems without the distraction of technical complexities.

DON'T BELIEVE THE HYPE: AGI IS FAR FROM INEVITABLE

Despite claims from large tech companies, achieving artificial general intelligence (AGI) with human-level cognition is still far away due to the need of both further improvements in AI along with more compute power.

Y COMBINATOR TRADED PRESTIGE FOR GROWTH

Y Combinator is experiencing a decline due to its overemphasis on growth over prestige.

Schedule a demo to learn more

OPENAI IS GROWING FAST AND BURNING THROUGH PILES OF MONEY

OpenAI's monthly revenue hit $300 million in August and it expects about $3.7 billion in annual sales this year. However, it expects to lose roughly $5 billion after paying for costs related to running its services. There will also be other costs such as equity-based compensation to employees. Its biggest cost is the computing power it gets through a partnership with Microsoft. OpenAI plans to raise ChatGPT Plus to $44 over the next five years.

META'S NEW AI-MADE POSTS OPEN A PANDORA'S BOX

Meta is starting to test personalized content created by Meta AI in Facebook and Instagram. Content will be generated based on interests, with some featuring the user's likeness. Users will be able to edit the content or swipe to see more content generated in real time. They will be able to opt out of the feature.

SPACEX LAUNCHES RESCUE MISSION TO RETURN STRANDED ASTRONAUTS

A SpaceX Falcon 9 rocket took off on Saturday with NASA astronaut Nick Hague and Russian cosmonaut Alexander Gorbunov on board. The mission has two empty seats to return Butch Wilmore and Suni Williams, who have been stranded for months on the International Space Station. Their stay was only supposed to last 8 days, but a problem with Boeing's Starliner's propulsion system prevented their return. Their total stay will end up lasting 8 months.

GOOGLE UNVEILS ALPHACHIP AI-ASSISTED CHIP DESIGN TECHNOLOGY — CHIP LAYOUT AS A GAME FOR A COMPUTER

Google's AlphaChip reinforcement learning method for designing chip layouts promises to substantially speed up the design of chip floorplans and make them more optimal in terms of performance, power, and area. The method has been instrumental in designing Google's Tensor Processing Units. Its success has inspired a wave of new research into using AI for different stages of chip design. The method has been shared with the public - a link to the repository is available in the article.

👉 Try Pluralsight for free

MITATA

mitata is a tool for benchmarking. It features automatic garbage collection and universal compatibility with engines and runtimes. mitata makes augmenting benchmarks easy and provides helpful warnings. It provides visualizations in the terminal as well as clear summaries.

IT'S HARD TO WRITE CODE FOR COMPUTERS, BUT IT'S EVEN HARDER TO WRITE CODE FOR HUMANS

Writing code that other humans interact with, like frameworks, libraries, APIs, SDKs, and programming languages, is difficult because developers need to grapple with other users' mental models of their code. It is equal parts computer science and psychology. This post discusses how to make these products easier for developers to adopt.

WHEN TO DO WHAT YOU LOVE

There's a lot of selection bias in advice about whether to follow your passion. Most people telling you to follow your passion became successful by doing what they're most interested in. One of the requirements for doing great work is that the person is working on what interests them most. Not everyone can do or wants to do great work. However, the root of great work is a sort of ambitious curiosity which can't be manufactured.

THE MESSY WORDPRESS DRAMA, EXPLAINED

WordPress co-founder Matt Mullenweg last week came out with a harsh attack on major WordPress hosting provider WP Engine, calling the company a 'cancer' to the community. This has escalated into a barrage of legal threats that has caught many website operators in a conflict beyond their control. WP Engine customers were off from accessing WordPress.org's servers, but the company has been granted a temporary reprieve. WP Engine claims that Automattic demanded a very large sum of money days before Mullenweg's attack and threatened to carry out a 'scorched earth nuclear approach' toward WP Engine if the amount wasn't received. Many more details about the situation are available in the article.

invest

BEING RAISED BY THE INTERNET

A childhood spent tinkering with computers has resulted in a successful career, largely thanks to the kindness of strangers who were willing to share their work openly.

KAMAL 2.0 RELEASED

Kamal, a deployment tool for running web apps directly on VMs or bare metal servers, is now simpler to use at any scale.

THE PUBLIC DOMAIN PROBLEM

A Creative Commons Zero license is the best way to guarantee that authors or their successors won't try to reclaim their work.

1 IN 6 COMPANIES ARE HESITANT TO HIRE RECENT COLLEGE GRADUATES

Gen Z has a reputation for being challenging to work with and difficult to manage.

DELL SALES TEAM TOLD TO RETURN TO OFFICE 5 DAYS A WEEK, STARTING MONDAY

Dell's new policy applies to salespeople worldwide - they were previously allowed to work remotely two days per week.

GOV. NEWSOM VETOES CALIFORNIA'S CONTROVERSIAL AI BILL, SB 1047

SB 1047 is a high-profile bill authored by State Senator Scott Wiener that would have made companies that develop AI models liable for implementing safety protocols to prevent 'critical harms'.

THE INTELLIGENCE AGE

AI is poised to amplify human capabilities, enabling tasks that seem unimaginable today. Through deep learning and increased computational resources, AI will offer advancements like personal assistants, educational tutors, and healthcare advisors. The focus must remain on making AI accessible and mitigating its risks, paving the way for widespread prosperity in the Intelligence Age.

APPLE NOT INVESTING IN OPENAI AFTER ALL, NEW REPORT SAYS

Apple has opted out of investing in OpenAI's $6.5 billion funding round, leaving Microsoft and Nvidia as potential investors. Microsoft is reportedly expanding its investment in OpenAI by another $1 billion. Despite not investing, Apple and OpenAI are partnering to integrate ChatGPT into iOS 18.

HOW ALPHACHIP TRANSFORMED COMPUTER CHIP DESIGN

AlphaChip is a reinforcement learning model that significantly accelerates and optimizes chip design processes. It generates superhuman chip layouts, currently used in Google's TPUs, in hours rather than weeks. This AI-driven method has broad applications, influencing both Google's hardware and external organizations like MediaTek.

A FASTER SUCCESSOR TO RNNS

The Parallel Gated Network (PGN) is a novel architecture designed to overcome the limitations of traditional RNNs in handling long-term dependencies. It reduces the information propagation path and uses gated mechanisms to effectively capture historical and current time step information.

FASTER IMAGE SUPER-RESOLUTION

DoSSR is a diffusion-based super-resolution model that enhances both performance and efficiency by leveraging pretrained diffusion models and starting the process with low-resolution images.

FASTER SAMPLING FOR FLOW-BASED MODELS

FlowTurbo is a technique designed to speed up sampling in flow-based models without sacrificing quality.

SOUND EVENT DETECTION

This repository introduces the Prototype-based Masked Audio Model, which improves sound event detection by making better use of unlabeled data. This technique generates pseudo labels using a Gaussian mixture model to guide the training of a Transformer-based audio model.

EFFICIENT LOW-BIT QUANTIZATION FOR LLMS

Vector Post-Training Quantization is a method designed to enable extremely low-bit quantization of large language models, optimizing memory and storage use during deployment.

FAST AND PHOTOREALISTIC HEAD AVATARS

LightAvatar is a head avatar model that improves rendering speed and efficiency using neural light fields (NeLFs).