Search the archive

AVANTE.NVIM

avante.nvim is a Neovim plugin that provides users with AI-driven code suggestions and the ability to apply these recommendations directly to their source files with minimal effort.

WHY YOU SHOULD USE CURSORS TO ITERATE OVER LARGE DATASETS IN REDIS

Managing large datasets in Redis can be tricky - using cursors with the SCAN command makes it much easier.

* Unlimited parallel test runs

* 24-hour maintenance and on-demand test creation * Zero flakes, guaranteed

Schedule a demo to learn more

BOEING STARLINER RETURNING EMPTY AS NASA TURNS TO SPACEX TO BRING ASTRONAUTS BACK FROM ISS

Boeing will return its Starliner capsule without the NASA astronauts that it delivered to orbit. The two astronauts will return via SpaceX's Dragon spacecraft on a mission launching on September 24. They will stay at the International Space Station for about six more months before flying home in February. The test was originally intended to last about nine days. This decision was made due to NASA's commitment to safety - Starliner's crew flight test faced problems, most notably with its propulsion system. NASA officials still support Boeing and Starliner will likely be able to launch with a crew again someday.

SHOCKER: FRENCH MAKE SURPRISE ARREST OF TELEGRAM FOUNDER AT PARIS AIRPORT

French authorities have detained Pavel Durov, the founder of the Telegram messaging app. They plan to hit him with charges related to abetting terrorism, fraud, money laundering, and crimes against children. There is a near-total lack of moderation on the platform, making it a top tool for organized crime. Durov's arrest has led to widespread panic in Russia - Telegram is used by Russian forces to communicate, and Western intelligence services may be able to obtain sensitive information about the Russian military if they gain access to it.

WORLD-FIRST LUNG CANCER VACCINE TRIALS LAUNCHED ACROSS SEVEN COUNTRIES

BioNTech has begun trials for BNT116, an mRNA lung cancer vaccine, in 34 research sites across the UK, US, Germany, Hungary, Poland, Spain, and Turkey. The vaccine works by presenting the immune system with tumor markers to prime the body to fight cancer cells expressing those markers. It aims to strengthen a person's immune response to cancer while leaving healthy cells untouched. The treatment is simple to deliver - patients receive a round of shots every week for six consecutive weeks and then every three weeks for 54 weeks.

CHINA'S OWN TESLA OPTIMUS? BEIJING'S AMBITIONS IN HUMANOID ROBOTS IN FULL DISPLAY AT EXPO

Chinese technology companies unveiled more than two dozen humanoid robots at the five-day 2024 World Robot Conference in Beijing last week. The show attracted 400 industry experts and academics and more than 160 domestic and overseas robotic firms. While there seems to be growing interest in robots from consumers, there are still several hurdles to overcome before their use can become widespread. These include a lack of sufficient data to train the AI used in the robots, limited use cases, and high prices.

BUFFDB

buffdb is a lightweight and high-performance persistence layer for gRPC. It is designed to be used in environments where bandwidth is limited and performance is critical. buffdb uses protobuff as its messaging format of choice but support for other formats is planned. It has full support for SQLite and partial support for DuckDB and RocksDB.

RUROCO

ruroco is a tool that lets users execute commands on a server by sending UDP packets. It contains several security features that prevent port scanning and clients running unauthorized commands. It encrypts data sent from client to server using RSA and it runs server software in a way that uses as little operating system rights as possible. ruroco can be used for single packet authorization, enabling webservices, and more.

WHY DOES GETTING A JOB IN TECH SUCK RIGHT NOW? (IS IT AI?!?)

A lot of new CS graduates are finding it really hard to get a job. There is a massive confluence of factors, both on the demand and supply side of the market, that are contributing to it being harder to land a job in tech. There are more CS graduates than ever and employers are hiring fewer software engineers. AI is not replacing engineers' jobs yet and it probably won't happen for an incredibly long time. This article looks at all these factors in detail.

MERCHANTS OF COMPLEXITY

It's hard to sell simple because it looks easy. Clever and sophisticated beats basic and straightforward most days in the market. These products imply there is something special, which allows them to command premium prices. Tech companies tend to turn basic problems people could easily solve for themselves quickly and cheaply into challenges that only a sophisticated solution can solve - usually with a subscription. People can escape this cycle by deciding that all the extra complexity isn't worth it and daring to be basic.

thank you. Free for 30 days for TLDR readers

HOW WE BUILT TOWNIE

Townie is an app that generates fullstack apps.

ELON MUSK SAYS X STAFF CAN GET THEIR STOCK — IF THEY PROVE THEY DESERVE IT

Staff at X will have to submit a one-page summary to leadership on their contributions to the company in order to get their stock.

JS DATES ARE ABOUT TO BE FIXED

One of the changes coming to ECMAScript is a native object specifically designed to represent dates and times with their corresponding time zone.

HOW ELON MUSK GOT TANGLED UP IN BLUE

Elon Musk thought Twitter Blue, which allowed die-hard users to pay for premium features, represented an untapped financial opportunity.

CREATE AN INTERNAL CLI

Some companies use their own internal CLIs for org-wide things like installing useful tools, generating boilerplate code, or running complex AWS commands - this article discusses how to get started with your own CLI.

CORS IS STUPID

CORS is a giant hack to mitigate legacy mistakes, but none of its protections are actually sufficient to solve the intended problem.

Cursor, Jasper, and many others. Start integrating today.

MICROSOFT RELEASES POWERFUL NEW PHI-3.5 MODELS, BEATING GOOGLE, OPENAI, AND MORE

Microsoft launched three new models in its Phi series: Phi-3.5-mini-instruct, Phi-3.5-MoE-instruct, and Phi-3.5-vision-instruct, each addressing different AI tasks with impressive benchmark performances. These models are open source under the MIT License and available for developers on Hugging Face. Despite their smaller size compared to some counterparts, the Phi models have achieved near-state-of-the-art results, surpassing competitors like GPT-4o and Llama in certain benchmarks.

WHAT'S REALLY GOING ON IN MACHINE LEARNING

Stephen Wolfram, the creator of Wolfram|Alpha, wrote a great exploration about different phenomena going on in machine learning models. He discovered a set of dynamics about the non-linearities in simple neural networks.

AGI SAFETY AND ALIGNMENT AT GOOGLE DEEPMIND

Google DeepMind's AGI Safety & Alignment team shared a detailed update on their work focused on existential risk from AI. Key areas include amplified oversight, frontier safety, and mechanistic interpretability, with ongoing efforts to refine their approach to technical AGI safety. They highlighted recent achievements, collaborations, and plans to address emerging challenges.

ENERGY-EFFICIENT AUTONOMOUS DRIVING WITH SPIKING NEURAL NETWORKS

Spiking Autonomous Driving (SAD) is the first unified Spiking Neural Network (SNN) aimed at addressing the energy challenges of autonomous driving.

EFFICIENT LANGUAGE MODEL TRAINING

Inheritune is a streamlined method for developing smaller base language models from larger ones by inheriting a few transformer blocks and training on a tiny fraction (0.1%) of the original data. This approach enabled the creation of a 1.5B parameter model using only an A6000 GPU in less than half a day, achieving results that compete well against larger models trained on significantly more data.

TIME SERIES REPRESENTATION LEARNING WITH DIFFUSION MODELS

Time Series Diffusion Embedding (TSDE) is the first diffusion-based approach for time series representation learning. TSDE segments time series data and uses dual-orthogonal Transformer encoders with a crossover mechanism to generate informative embeddings.

EFFICIENT TRITON KERNELS FOR LLM TRAINING

The Liger Kernel, released surprisingly by LinkedIn, is an efficient set of kernels for language model training. It increases throughput by 20% and decreases memory usage by almost 60% for the popular Llama models. Importantly for practitioners, it only requires 3 lines of code change and integrates with many popular modeling frameworks.

POSTGRESQL IS A GOOD VECTOR DATABASE

pgvectorscale builds on pgvector with higher performance embedding search and cost-efficient storage for AI applications. It is around 28x faster than competitive and commonly used vector stores.

REDUCING GENDER BIAS IN LLMS

GenderCARE is a comprehensive framework aimed at assessing and mitigating gender biases. It introduces innovative criteria for evaluating gender bias, emphasizing inclusivity, diversity, and objectivity.

TEACHING CHAT MODELS TO SOLVE CHESS

Traditional base language models are quite good at playing chess, averaging around 1800 elo. However, chat models tend to struggle with a precipitous drop in performance. This post explores how to tune chat models like GPT-4o to play chess through a combination of prompting and fine-tuning.

DATA EXFILTRATION FROM SLACK AI VIA INDIRECT PROMPT INJECTION

A vulnerability in Slack AI was discovered where attackers can exfiltrate data from private channels they aren't members of via indirect prompt injection. It allows attackers to use public channel messages to manipulate the LLM into divulging private information, such as API keys, in query responses. This issue, alongside a phishing attack vector, persists even after Slack AI's August 14th update incorporating channel and DM files, significantly expanding the risk surface area for such exploits.

GENERALIZED SEGMENT ANYTHING MODEL

Generalized SAM (GSAM) is a new method for fine-tuning the Segment Anything Model (SAM) with variable-size images more efficiently.

NEW SIGLIP MODEL (HUGGING FACE HUB)

Google has released a new SigLIP model that uses a shape optimized vision transformer model architecture.

3D OCCUPANCY ESTIMATION

GaussianOcc is a method for fully self-supervised and efficient 3D occupancy estimation using surround views.

BRINGING LLAMA 3 TO LIFE

Meta has introduced Llama 3.1, an advanced open-source LLM that brings new capabilities like synthetic data generation and model distillation.

TELEGRAM FOUNDER AND CEO PAVEL DUROV ARRESTED IN FRANCE BY NATIONAL ANTI-FRAUD OFFICE

Pavel Durov, founder and CEO of Telegram, was arrested at an airport in France by the country's National Anti-Fraud Office. He was allegedly put into custody for offenses related to terrorism, drugs, fraud, money laundering, and other charges related to individuals' activity on the messaging app.

COURT RULES THAT NONE OF THE TOKENS TRADING ON KRAKEN ARE SECURITIES

The Federal Court for the Northern District of California ruled that none of the tokens on Kraken are securities and that the definition of SEC's “crypto asset security” is confusing and unclear. The case is proceeding to discovery, as agreements or contracts around a token could be considered a security. The SEC will have to prove that transactions on exchanges satisfy the Howey test, which is an arduous task to take across the industry.

SONY, ELECTRONICS PIONEER BEHIND WALKMAN, STARTS OWN BLOCKCHAIN 'SONEIUM'

Prominent Japanese electronics company Sony is launching its own Ethereum-based Layer 2, called Soneium. It is based on OP Stack technology and will be part of the Superchain Interoperability effort. The project aims to onboard existing Web3 users, and then integrate Sony products like Sony Bank, Music, and Pictures within two years.

NIGHTSHADE 2.0: NEAR'S NEW SHARDING DESIGN

Nightshade 2.0 allows for stateless validation, meaning NEAR validators no longer have to store their shard's state to verify state transitions and blockchain updates. It also improves performance and adds capacity for more shards - NEAR currently has six and plans to add four more by the end of the year. In 2025, the team will focus on dynamic re-sharding, which adjusts the number of shards based on network load and is considered “the holy grail of sharding.”

Don't miss out→

HYPERLIQUID: INNOVATING ON-CHAIN PERPETUAL TRADING

Hyperliquid has quickly become a leading DeFi platform, offering a high-performance L1 with an onchain perpetual exchange and spot trading platform. It has already surpassed established competitors in TVL and daily trading volume, introducing features like the Hyperliquidity Provider Vault for passive LPing. Its competitive fee structure, upcoming EVM support, and potential token release position it favorably, with valuations ranging from $1 billion to $3.5 billion.

HOW DOES THE ETHEREUM FOUNDATION SPEND ITS ETH?

The Ethereum Foundation holds $845 million in ETH, representing 0.25% of the total supply. It allocates funding to education, research, and development, which has led to the development of zero-knowledge cryptography and funded conferences, online courses, and novel projects like the ZK Email Wallet and Daimo payment protocol. The foundation recently came under fire for transferring $100 million to Kraken for future expenses as some in the community call for greater transparency.

FIRST-CLASS CRYPTO APPS

Polymarket is leading a new era of global-first applications by aggregating global liquidity with stablecoins, creating a universal market accessible from day one. This will transform the traditionally slow and costly process of scaling because of fiat limitations.

CRYPTO BULL RUN DRIVEN BY MACRO FACTORS

The current crypto bull run seems to be influenced by external macroeconomic factors. However, additional market boosts may stem from exciting upcoming projects and developments across various crypto platforms. Some of these include new mainnets like Monad and MegaETH, innovative token models like those proposed by Eigenlayer, significant upgrades on Ethereum (e.g., Pectra), and major rebranding initiatives like Fantom's transformation into Sonic.