Full-text search across article titles and summaries. Refine with sector, year, content type, and reading-time filters.
Most of the interesting conversations in tech now happen in private group chats. This thread looks at what makes a group chat good. The best groups are small, have good moderators, and share rituals, jokes, and routines. New ideas, combined with a sense of humor, is what makes great group chats work - boredom breaks them.
Microsoft is releasing the next Call of Duty to its Game Pass subscription service. Players will still have the option to purchase the game outright. While this plan may mean people could end up paying Microsoft less to play the new game, the company hopes that it will draw new users to Game Pass who will end up paying more over the long term. Starfield, which was released on Game Pass at launch last year, drove a record number of new subscriptions in a single day.
Sponsored
Master math, data, CS, and AI concepts in 15 minutes a day with Brilliant — guided interactive problem solving that's effective and fun. Your brain will thank you. Free for 30 days, then 20% off for TLDR readers [14]
AI PCs have neural processing units designed to specifically handle artificial intelligence tasks.
This article looks at Starlink and how TCP interacts with the characteristics of the Starlink service.
While email works pretty well for its users, the backend for email is a mess.
A provision in OpenAI's off-boarding papers states that employees risk losing all vested equity they earned during their time at the company if they don't sign a nondisclosure and non-disparagement agreement when they leave.
38% of web pages that existed in 2013 are no longer accessible.
Google and Meta have made it so GenAI is now free and ad-supported, forcing OpenAI to compete in search just to stay relevant.
→ See how you can simplify investigations and accelerate remediation
Sponsored
While some are worried about the use of AI in cyberattacks, Orca
shows how AI-powered security enables companies to:
xAI has released the weights and architecture for Grok-1. Grok-1 is a Mixture-of-Experts model with 314 billion parameters. The released model is a raw base model checkpoint from the Grok-1 pretraining phase, which means that it is not fine-tuned for any specific application. It was trained from scratch by xAI using a custom training stack.
Unitree has released the first few details on its G1 Humanoid Agent, a $16,000 robotic helper. The robot is equipped with 3D LiDAR sensors, a depth camera, three-fingered grippers, and a 9,000-mAh battery pack. It can recover from brutal kicks and punches, as seen in the short demo video in the article. At around 35 kg, the robot can fold itself down to 690x450x300-mm proportions for compact carry.
LoRA is a popular way to fine-tune models to inject style or knowledge. This paper explores the trade off between power and forgetting when using LoRAs. The research finds that LoRAs are able to learn less than full fine tuning, but maintain more of the original “out of distribution” performance.
Meta has announced a natively multi-modal model called Chameleon that operates on images and text simultaneously like GPT-4o. It outperforms many alternative models. The Meta team has since far surpassed this work on internal models.
Google's most recent family of models has an updated technical report. There is little information about the models or the data used, but there is significant information about the evaluation and safety measures taken, which is an interesting window into large-scale alignment.
Google has released a model explorer that makes it easy to inspect the computation graph of your models. It can be useful for debugging and performance engineering.
Using GPT-Fast, you can run the most recent Google VLM in real time. This opens up a number of powerful downstream tasks as the model is extremely easy to fine-tune for specific tasks.
Experts is the easiest way to create and deploy OpenAI's Assistants and link them together as Tools to create a Panel of Experts system with expanded memory and attention to detail.
Google kicked off its I/O developer conference with a two-hour presentation containing a rapid-fire stream of announcements. This article summarizes these announcements into an easy-to-read list. The announcements include Firebase Genkit, an addition to the Firebase platform aimed at making it easier for developers to build AI-powered applications in JavaScript/TypeScript; LearnLM, a family of generative AI models fine-tuned for learning; AI-generated quizzes on YouTube; a new discovery feature for apps in Google Play; scam call detection; Ask Photos, a feature that lets users search across their Google Photos collection using natural language queries; and much more.
Gamma raised $12M from Accel to reimagine presentations using AI. So far, 60M+ Gammas (AI-generated slides) have been created across 18M+ users.
Garple is a website that sells short, easy-to-say .com domains that would make good startup names. All the code was written by an AI coding agent.
Sponsored
Whether you're new to AI or a seasoned expert, this free webinar by MindStudio [18] will show you how to automate the tedious parts of
Julius AI has launched the first-ever AI code interpreter for the programming language R, potentially revolutionizing how academic researchers perform statistical analysis.
Find ideal business customers using AI and Bayesian models for optimal delivery and conversion rates.
An experiment in running the small MoonDream VLM directly on the web with WebGPU inference.
There are many complaints about new coins launching at the highest possible fully diluted valuation and immediately listing on a tier-1 CEX, providing exit liquidity for VCs. Over 80% of tokens listed on Binance in the past six months are down, with the only exceptions being two Solana ecosystem coins and two memecoins. The average FDV of a Binance listing is $4.2 billion. These projects have no organic community or positive retail sentiment, causing their tokens to perform poorly.
This thread scrutinizes the significant downturn of "low float, high FDV" cryptocurrencies listed on Binance, debunking prevalent theories about market manipulation by VCs or shifts in retail investor interest towards meme coins. None of the popular narratives convincingly explain the broad price declines. A combination of factors in a complex market environment, rather than a single culprit, is responsible for the observed trends. Free markets will eventually adjust these mispricings on their own.
Ethereum's transition to Verkle trees, a more efficient data structure, will provide various scalability and node operation benefits. It will allow for a 10x increase in mainnet gas, simplify the node architecture, and reduce node sync time to a few seconds. Consequently, the path for smartphone-based and light nodes will become possible, helping to improve Ethereum decentralization.
PureDai is a proposed stablecoin that aims to deliver cypherpunk ideals. It prioritizes extreme decentralization and resilience by leveraging decentralized collateral, maximally decentralized oracles, and minimal governance. The stablecoin will have a free-floating target price, permanent emissions to incentivize supply-side growth, and a governance token. It will interact with Maker SubDAOs, particularly the Spark lending protocol, to ensure efficient markets. It is not expected to be released for several years and will likely change significantly before then.
Ethereum has underperformed in 2024, with many questioning its role when L2s and competing L1s like Solana take market share. Coinbase still believes in Ethereum's long-term potential and its advantages surrounding the developer ecosystem, the proliferation of the EVM, ETH's utility as DeFi collateral, and its decentralization. ETH is uniquely positioned as a store-of-value and technology token and could outperform in the second half of 2024.
The liquid restaking token sector continues to grow, but ETH has been down since the EIGEN airdrop, leading to speculation that ETH was overbought or leveraged to farm the token. EtherFi and Renzo are leaders in the LRT market, with a 73% market share. Still, Renzo's ezETH has been trading at a 2-3% discount since a depeg event a few weeks ago, making Puffer's pufETH the second-most liquid LRT behind eETH. 56% of EigenLayer's TVL is used in a yield generation protocol.
A report from Binance Research reveals a substantial shift in the crypto market, with the circulating supply of tokens at launch decreasing significantly over the past three years, from about 41% to 12.7%. With $155 billion in token supply scheduled to unlock from 2024 to 2030, the study underscores the challenges and potential volatility facing projects with low float and high fully diluted valuations, highlighting the need for real utility to support such high initial valuations.
Vitalik Buterin reflects on Ethereum's progress and the need to prioritize decentralization and permissionlessness as the network scales. He mentions concerns about MEV, liquid staking, and node hardware requirements and suggests potential solutions like inclusion lists, reduced withdrawal periods for stakers, single-slot finality, and peer-to-peer networks for storing historical data. While Ethereum has moved away from its 2021-era ideas about offloading responsibilities to centralized actors, there is still room for improvement to make Ethereum a strong base layer for L2s.
This thread criticizes the reliance on Layer 2 (L2) solutions for blockchain scaling, arguing that it leads to worse user experience due to ecosystem fragmentation, poor economics, and encourages centralization, which compromises trust and security. It advocates for scaling at the Layer 1 (L1) level, contending that L2 solutions often remain centralized, introduce significant security risks including potential theft by administrators, and ultimately do not solve the inherent problems but merely shift them onto a weaker model.
Lido's co-founders and Paradigm are secretly backing a new "restaking" company, Symbiotic, poised to challenge EigenLayer and potentially reshape the decentralized finance (DeFi) landscape.
The Runes protocol launch experienced a "sell-the-news" reaction, but the community remains robust and development continues, with potential catalysts such as listings on major exchanges like Kraken and Binance poised to rekindle interest.
The Venezuelan government claims this measure will stabilize the national power supply, which has suffered frequent blackouts since 2019.
Eigenlayer enables users to leverage their staked ETH to obtain additional staking rewards while supporting new Actively Validated Services (AVS).
Rollups scale better than fast L1s due to a smaller replication factor, reducing communication overhead and enabling faster performance.
Sponsored
Fortune 50 companies like AWS and Google Cloud and fast growing startups like Vercel and Secureframe work with TLDR to reach our audience of over 4 million tech-savvy decision makers with highly customized campaigns (like this one!).
Samsung capitalized on Apple's controversial iPad Pro ad with a new campaign for the Galaxy Tab S9 Ultra. Apple's ad, which featured a hydraulic press crushing artistic tools, sparked debate that made the company decide not to run it on TV. In Samsung's response, a woman picks up a visibly damaged guitar at a location reminiscent of the Apple commercial and begins creating music using her tablet as a creativity ally.
Whatsapp's new layout isn't a radical departure - search is more prominent at the top of the app. Below it, tabs let you quickly access unread messages and groups, and on the bottom, a row of icons lets users access Chats, Updates, Communities, and other features. However, WhatsApp users on iOS have noticed a significant visual change: the predominant color accent on everything is now green instead of blue.
Despite its impressive capabilities, ChatGPT lacks UX standards and has limited user guidance, unclear error messages, a lack of context awareness, inconsistent output quality, and reduced transparency and explainability. Solutions to address these issues include a clearer and more intuitive user interface, better onboarding, integration of fact-checking mechanisms, and exploring ways to incorporate context awareness into prompting. Ensuring transparency and emphasizing user control over the generated content are also essential for trust.
Apple's iOS 18 introduces AI-driven accessibility features, including eye-tracking for device control, motion sickness reduction, and speech recognition for atypical speech patterns. The Vision Pro headset will also provide live captions for real-life conversations, demonstrating Apple's commitment to inclusive design for a broader range of conditions.