Search the archive

WEB-CHECK

Get an insight into the inner-workings of a given website: uncover potential attack vectors, analyze server architecture, view security configurations, and learn what technologies a site is using.

KUBEFUZZ

kubefuzz is a generative and mutative fuzzer for liberating admissions controller chains that works by parsing the cluster API specification. Users can supply constraints to guide the fuzzer’s resource generation.

BLACK HAT USA 2023 SLIDES

Here are all the slides presented at Black Hat USA 2023.

INTRODUCING THE FIRST AWS SECURITY HEROES

The AWS Heroes program recognizes community members that have combined technical expertise and a desire to help the community. AWS has now created a specialized category for Security Heroes.

AI.COM FLIPS FROM CHATGPT TO ELON MUSK’S X.AI

AI.com now redirects to x.AI, a company that aims to understand the true nature of the universe. OpenAI had seemingly purchased this domain back in February, but it looks like Musk has swooped in.

THE RIGHT TO PAY ANONYMOUSLY HAS BECOME PART OF AN EU CULTURE WAR

There’s a growing movement across the EU to save cash as governments, banks, and merchants encourage consumers to go digital.

Serve Robotics, an autonomous sidewalk delivery robot startup that spun out of Uber’s acquisition of Postmates, is going public via a reverse merger with a blank-check company.

SHOULD LATER STAGE STARTUPS CONSIDER AN ANNUAL LIQUIDITY PROGRAM?

Liquidity puts capital into the hands of employees and early investors, enabling them to reinvest into the ecosystem or themselves. We still see very few secondary events in the startup world, but it looks like that might change, with all the new products and communities propping up to support private share sales.

THE PRINCIPLES OF GREAT STRATEGY

A lot of people have a romantic view on strategy. You do this big lift, lock in a strategy, and then flip to pure execution. In reality, building a company is this dance between strategy and execution. Follow these principles to avoid writing a strategy doc that nobody ever looks at more than once.

PLAID: FINANCE’S NEXT GREAT NETWORK

A long deep dive into Plaid and how its product, model, and culture led it to become a FinTech giant.

GOOGLE’S PROJECT IDX

An AI-enabled browser-based development environment for building full-stack web and multiplatform apps

“TECH COMPANIES”?

A zero interest-rate policy environment can drive a lot of things in the wrong direction. This article unpacks some ‘winner takes none’ outcomes, where even the best/surviving companies end up consuming a tremendous amount of dollars and delivering little value.

REAL-WORLD DIVERSIFICATION

As a big advocate of side-gigging and combinatorial creativity, I like this approach of using hobbies and extracurricular activities to build out T-shaped knowledge. What you do outside of work is going to influence how you show up to work.

THE RACE TO CAPTURE VALUE: CLOUD LESSONS FOR THE AI ERA

We get it, AI is changing everything. But with any disruptive technology, doing the math on that disruption is actually pretty hard. Enter a16z with their typical deep-dive on how AI is going to disrupt the cloud and application layers. Dense (although good) read.

HUMAN CREATIVITY AMONGST AI DOMINANCE

More marketers and content producers are turning to AI for efficiency. Over time, this will lead to a level of genericism in the market. The antidote will be authentic human production.

CARROT PROBLEMS

Advice often comes in the form of someone giving you their winning lottery ticket numbers, but that’s not exactly helpful. It’s good to be mindful when you hear successful people explain some of their success, that it might not always be the truth. Red herrings can lead you down costly paths.

TOP 10 SAASTR TIPS FOR FOUNDERS & CEOS

We’re big fans of the folks over at SaaStr and this recap of a bunch of their tips for founders & CEOs resonates this time of year when we’re taking a break and also preparing for the big push into Q4. Always insightful and timely advice.

SCALING THE INSTAGRAM EXPLORE RECOMMENDATIONS SYSTEM

Explore is a recommendation system on Instagram that uses machine learning to show users content they'll like. Meta uses advanced machine learning models, like Two Towers neural networks, to help Explore be more scalable and flexible. This article goes over the techniques the Meta team uses, such as ranking and Bayesian optimization.

EXPLORING THE INTERNALS OF LINUX V0.01

The Linux kernel is often mentioned as an overwhelmingly large piece of open source software. However, the first version of Linux, v0.01 was pretty small. It consisted of only 10,239 lines of code. This article dives through some interesting tidbits in Linux v0.01, such as system calls and what the file system supported at the time.

EVERY PHONE SHOULD BE ABLE TO RUN A PERSONAL WEBSITE

Back in 2009, anyone with a Nokia phone could have a personal website running on their own phone. However, this was never widely adopted because there was no incentive for large companies to standardize it. Being able to run a server on your phone should be more normal, especially with the computing power that phones today are able to provide.

SPEEDRUNNING GUIDE: MID-LEVEL (L4) -> SENIOR (L5)

Senior engineer (L5) is the first level that big tech recognizes as a “terminal” level. This article gives actionable advice on how to get from being a mid-level to a senior engineer as fast as possible. The advice includes taking on impactful projects from the start, focusing on team leadership and influence, and uplifting one’s teammates.

INFISICAL

Infisical is an open source, end-to-end encrypted secret management platform that teams can use to centralize their secrets like API keys, database credentials, and configurations. It features a user-friendly dashboard, client SDKs, a CLI, and more.

CODELITE

CodeLite is an open source, free, cross-platform IDE, specialized in C, C++, Rust, Python, PHP, and JavaScript, mainly for backend developers using Node.js.

THE CARROT PROBLEM

During World War II, the British invented radar to help their pilots shoot down German planes at night. To keep this technology a secret, they spread propaganda that their pilots had developed exceptional eyesight from eating carrots. This is an example of a "Carrot Problem," where people are misled into wasting time or effort on something that doesn't work.

WAIT, CHATGPT DIDN'T TAKE MY JOB?

Artificial intelligence is often portrayed as a job killer. However, instead of cutting jobs, companies are using the technology to get more done with the same number of people. For example, Allen & Overy, a firm that employs more than 3,000 lawyers worldwide, started working with a generative AI tool called Harvey last year and hasn’t replaced a single person with it. Harvey helps employees perform better and adds value to the firm.

OF EXECUTIVES HAVE EGG ON THEIR FACES'

80% of bosses regret their initial return-to-office decisions and say they would have approached their plans differently if they had a better understanding of what their employees wanted. Many companies are realizing they could have been a lot more measured in their approaches by focusing on employee data first and foremost.

20 YEARS OF VIM

This chart shows the last 20 years of development in the Vim git repository.

REACT: WHAT’S WRONG WITH CONDITIONAL RENDERING?

3 ways to spot conditional rendering patterns in React and how you can improve them.

A MOTHER-DAUGHTER DUO

CEO CLAIMS X, FORMERLY TWITTER, IS CLOSE TO ‘BREAK EVEN’

EXPERIMENTAL INSULIN IMPLANT USES ELECTRICITY TO CONTROL GENES

LUNAR LANDING SINCE 1976

RETAKE

Retake is a hybrid search for Postgres. It accelerates the development of AI applications that require searching, matching, or analyzing database documents by connecting directly to Postgres. Retake can manage separate vector stores and text search engines, upload and embed documents, and reindex data. It ensures that indexed data is always kept in sync and it updates the embedding/vector representation of data behind the scenes whenever data is changed.

THINGS YOU FORGOT (OR NEVER KNEW) BECAUSE OF REACT

MORE REAL

ARE REPORTS OF STACKOVERFLOW’S FALL GREATLY EXAGGERATED?

COMPUTING

THE FUTURE OF TECHNOLOGY LOOKS A LOT LIKE...PIXAR?

A SENIOR ENGINEER/EM JOB SEARCH STORY

CRACKDOWN COMING

RESEARCHERS WATCHED 100 HOURS OF HACKERS HACKING HONEYPOT COMPUTERS

PGAUDIT

This PostgreSQL extension provides detailed session and/or object audit logging via the standard PostgreSQL logging facility.

STANFORD AGENTS CODE OPEN SOURCED!

A research group at Stanford created a splash a few months ago with their fully simulated village with realistic people living lives. That code is now open sourced. The most interesting piece is likely how they prompt engineered each person and the overall system to work with a non-state preserving API like gpt-3.5-turbo.

CHATGPT EXPANDS ITS ‘CUSTOM INSTRUCTIONS’ FEATURE TO FREE USERS

OpenAI announced that it’s expanding custom instructions — a way to give users more control over how ChatGPT responds — to all users, including those on the free tier of the service.

NEWS OUTLETS DEMAND NEW RULES FOR AI TRAINING DATA

Several media organizations have called for rules protecting copyright in data used to train generative AI models. They have signed an open letter urging lawmakers around the globe to consider regulations that require transparency in training datasets and the consent of rights holders before using data for training. The letter also requests allowing media companies to negotiate with AI model operators and identify AI-generated content, and mandates that AI companies eliminate bias and misinformation in their services.

ENHANCING LLAMA-2 MODEL'S PERFORMANCE THROUGH IN-CONTEXT LEARNING

This paper studied in-context learning with the pretrained Llama-2 language model by retrieving 9 demonstration examples during chat-style prompts. The method, which didn't alter the model's weights, resulted in a 7x win-rate increase compared to the text-davinci-003 model, making it on par with models that underwent alignment fine-tuning.

IMPROVING 3D OBJECT DETECTION USING DETR

Current 3D object detectors often misinterpret data, causing them to focus on distant points. The authors of this paper have developed a new method, 3D Vertex Relative Position Encoding, that directs the detector's attention to points close to the target objects.