75% of the most visited websites in the US and Europe are noncompliant with major privacy regulations like the CPRA and GDPR. In the US, top websites share personal data with an average of 17 third-party advertisers. In Europe, 6.79% of media and ecommerce sites fail to meet privacy standards. Since 2022, at least 10 US companies have been fined for noncompliance, and Amazon was fined $888 million in Europe for improper targeting.
The "3-second rule" for marketing video ads says that you should address the audience within the first 3 seconds by mentioning their job title, problem, or group. This tactic taps into the brain's focus on identity-related information, boosting engagement. Companies like Deel, Gong, and Hubspot use it to target roles like HR Managers, Sales VPs, and Marketers. It works well on broad-reach channels like YouTube or TV, where audience targeting is less precise.
Brand engagement on Threads is thriving because the platform allows social media managers to post relatable, spontaneous content without heavy oversight or approvals. Brands like TSA, USPS, Beyond Meat, Calm, and Notion are thriving by embracing this freedom, posting at all hours with casual, unfiltered updates.
Google reversed its decision to deprecate the popular Chrome extension, Tag Assistant, after user feedback. This move consolidates Tag Assistant Legacy and Tag Assistant Companion into a single, more efficient tool.
Faceless videos use voice overs, animations, and images to promote products or share information. They're cost-effective as they require less expertise and can be easily produced with AI tools. Tools like Animoto, Jasper, Copy.ai, Grok, ElevenLabs, and Murf.ai help streamline idea generation, scripting, visuals, voiceovers, and editing.
Brands will be leaning more into unhinged content this year to connect authentically. IRL events will bring digital communities into real-world experiences. Social commerce will keep transforming shopping habits, simplifying the path to purchase. Cultural participation and community-driven campaigns will play key roles in reshaping engagement, in-person connections, and how audiences interact with brands.
Birkenstock's 2023 sales rose by 20% following a rebranding that segmented audiences into 4 core groups: comfort seekers, independents, acceptance seekers, and balancers. The brand revamped its architecture, phased out underperforming sub-brands, and emphasized its core value of “genuine spirit” with a focus on craftsmanship and originality. The overhaul also included modernized retail experiences, edgier visual merchandising, and immersive pop-up events.
Households with GLP-1 drug users, like Ozempic, reduced grocery spending by 6% within 6 months. Spending at fast-food restaurants also declined, particularly at breakfast and dinner. The average household spends $416 less annually, with reductions in high-calorie, high-sugar foods. In response, chains like Smoothie King have adapted by creating GLP-1 Support Menus. GLP-1 usage, which has surged 32% since June, is reshaping consumer dining habits and food spending.
A content framework that focuses on 3 goals, buyer profiles, themes, channels, and metrics.
A breakdown of what drives success at each step of Jones Road's $150M customer funnel.
Reducing ad fatigue involves refreshing creatives, rotating campaigns to target new segments, and experimenting with frequency caps every 2-3 weeks.
A Mozilla engineer noticed that poorly-behaved bots, primarily from OpenAI and Amazon, caused 70% of the load on Diapsora's servers (an open-source social network). To address this, he built a “polite and fast” web crawler using Python and gevent, prioritizing respectful rate limiting by defaulting to one request per domain every five seconds and dynamically adjusting based on server responses. The crawler efficiently manages the queue by fetching robots.txt files in parallel and uses a sophisticated refetching strategy using HTTP headers to minimize redundant requests.
ComfyDeploy migrated its dashboard from Next.js to a plain React setup, which improved build times (from 3 minutes to 18 seconds) and hot reload speed (under 200ms). This decision was driven by escalating Vercel costs, testing difficulties with Next.js's server actions, and painfully slow local development. The migration incorporated TanStack Router and Rspack, forcing architectural improvements and cleaner code.
Inferable chose HTTP long polling over WebSockets for its real-time system due to scalability challenges with its Node.js/TypeScript backend and PostgreSQL database. Long polling, unlike short polling or WebSockets, keeps the connection open until data is available or a timeout is reached, offering a balance between immediacy and resource efficiency. This approach avoided complexities associated with WebSockets, such as adapting its observability stack, authentication mechanisms, and infrastructure configurations.
There are two developer types: "bootstrappers," who use AI for rapid prototyping, and "iterators," who use AI in daily workflows. While AI accelerates development a lot, it introduces challenges like the "70% problem," where achieving the final 30% of software quality requires real human expertise. As a result, AI makes the lives of experienced software engineers easier, but doesn't replace them.
2025.01.06·newsletter.pragmaticengineer.com·25 min read·TLDR Web Dev Sam Altman reflects on OpenAI's journey, the unexpected success of ChatGPT's launch, and the subsequent rapid growth. It was challenging and messy to build a company around groundbreaking technology, as shown by his own public firing and subsequent reinstatement. He's learned a lot and will be more confident moving forward.
Sponsored
Need to add WYSIWYG HTML editing to an app or a website? CKEditor is a modern, feature-rich JavaScript editor that's infinitely customizable [9] (frontend + backend) to build exactly what you need. It's production-ready, comes with a sleek and collaborative UX, and works with your tech - TypeScript, React, Vue.js, and more. Start for
LogLayer is a JavaScript logging library that provides a consistent logging experience across different logging providers. It offers a flexible API for adding tags, metadata, and errors to log messages, supporting multiple loggers simultaneously (e.g., DataDog and New Relic).
A Tailwind CSS plugin that lets you animate any element with inline styles. It also comes with a Chrome extension to let you play with animating your website directly.
Experiment with different flex properties to understand how they affect layout. CSS code is generated in real time and controls are available for finetuned adjustments.
This developer extracted an AI model from Microsoft's Seeing AI app. The app's model was found encrypted within the APK file as a TensorFlow Lite model. It was then verified using Netron, which led to the successful extraction of the AI model's weights and biases.
An LLM was told to "write better code," which led to performance improvements in a Python program through multiple iterations. However, the improvements eventually plateaued, and the code became over-engineered. A second experiment used more precise prompt engineering, explicitly outlining optimization goals, which eventually led to a 100x speedup.
The CSS-Tricks website returned to consistent publishing in 2024, leading to 26 million unique pageviews, many significant site updates and content additions, and new goals.
This author gained remote code execution on a Yamaha PSR-E433 synthesizer via its MIDI interface by reverse engineering its firmware.
Devs should take a pragmatic approach to debugging, and all techniques are valid (including just using printf) for efficiently finding and fixing bugs.
After both maintainers of Recoil, a React state management library created by Meta, were laid off, the repository was officially archived as no Meta engineers were willing to take up the work of maintaining it.
Tesla shares fell by around 7% in trading on Thursday after Tesla released its fourth-quarter vehicle production and deliveries report. The company's shares finished up 63% for the year. They reached an all-time high in mid-December. Tesla aims to offer lower-cost and autonomous vehicles in 2025. This is predicted to lead to 20% to 30% growth over 2024.
SpaceX's Starlink now serves over 4.6 million users across 118 countries and territories. The company reported 3 million users back in May. It has since expanded its services to 27 new markets. SpaceX completed 89 launches dedicated to building and enhancing the Starlink satellite constellation throughout last year. The Starlink network currently has over 7,000 active satellites in low Earth orbit, with nearly 350 designed to connect directly to LTE phones.
2025.01.03·techstory.in·2 min read·TLDR Anthrobots are synthetic biological proto-organisms made of living cells with un-edited genomes. They could have many future uses in the human body, including laying down pro-regenerative molecules, clearing plaque from arteries, healing spinal cord or retinal damage, dealing with cancer cells or bacteria in the gut, or informing doctors of the status of the surrounding tissues. Anthrobots do not exponentially reproduce and just biodegrade at the end of their lifespan. Scientists are still experimenting with these lifeforms to get a true idea of their capabilities in various problem spaces.
2025.01.03·links.tldrnewsletter.com·17 min read·TLDR Wi-Fi 8 looks to be focused on greater reliability rather than on pushing bandwidth even higher. It will make the speed gains promised by Wi-Fi 7 more of a reality. End users will not see the full benefits until they have upgraded both their access points and endpoint devices to the new standard. The first devices with Wi-Fi 8 are not expected to roll out until the second half of 2028.
2025.01.03·theregister.com·7 min read·TLDR Zasper is an IDE designed to support massive concurrency. It has a minimal memory footprint and provides exceptional speed. Zasper can be used to run REPL-style data applications like Jupyter notebooks. It is currently fully supported on Mac, with limited support on Linux.
2025.01.03·github.com·GitHub repository·TLDR Deepseek's flagship model, v3, is a 607B Mixture-of-Experts model with 37B active parameters. The company achieved remarkable performance while keeping training costs surprisingly low. The model outperforms other alternatives at its price point. It excels at reasoning and math, surpassing GPT-4 and Claude 3.5 Sonnet.
2025.01.03·composio.dev·10 min read·TLDR Heaps of companies want employees with long-term experience. Many of these companies aren't attractive to younger workers and would be glad to hire someone with experience. Experienced workers can also enter into consulting/professional services, which provides a varied and interesting work environment. The expiry date for a software engineer is definitely much later than 40 - many are still working late into their 50s and 60s.
2025.01.03·news.ycombinator.com·TLDR Many people are convinced that companies are targeting ads to them by listening through their phone's microphones. For this to be true, companies would have to record audio snippets, transmit them back to their servers for additional processing, and then feed that data into a system that forwards it to advertising partners. This is far-fetched, especially for companies like Apple, which has a brand and reputation as a privacy-first company and spends large amounts on product design and engineering to prevent apps from doing exactly this kind of thing. The more likely scenario is that the quality of targeting is high and that people just aren't that different from each other.
2025.01.03·simonwillison.net·4 min read·TLDR The Sixth Circuit Court of Appeals said that the FCC lacked the authority to oversee wireless and home-broadband services.
2025.01.03·links.tldrnewsletter.com·2 min read·TLDR Object identifiers are a standardized key-value system for naming any object or concept.
2025.01.03·progress.compose.sh·10 min read·TLDR If approved, Apple will pay up to $20 per Siri-enabled device (for up to five devices) to customers who made purchases between September 17, 2014, and December 31, 2024.
2025.01.03·arstechnica.com·3 min read·TLDR Last week, AWS added another layer to S3 Buckets and introduced S3 Tables for storing data in Apache Iceberg.
2025.01.03·theregister.com·3 min read·TLDR A reflection of how life has been after selling a company from Vinay Hiremath, the CTO and co-founder of Loom.
The projects include 'This is Not My Name' by Liuhuaying Yang, 'Your Name In Landsat' by NASA Landsat Science, and the 'Climate—Conflict—Vulnerability Index' by the German Federal Foreign Office.
2025.01.03·flowingdata.com·4 min read·TLDR OpenAI faced criticism for not delivering the promised opt-out tool by its deadline, raising concerns over user data practices and transparency.
2025.01.03·techcrunch.com·4 min read·TLDR AI Google announced third-party tool support for Gemini Code Assist, enhancing its AI code completion capabilities. The tools, available in private preview, connect Code Assist to external applications, increasing productivity and minimizing context switching. Initially, tools are limited to Google Cloud partners, with offerings from GitLab, GitHub, and others at launch.
2025.01.03·techcrunch.com·3 min read·TLDR AI A fascinating paper that explores closed form equations that can model generated images from diffusion models. This means that with a high degree of confidence you can predict the image that will be generated, in a simpler setting.
2025.01.03·arxiv.org·21 min read·TLDR AI Researchers have trained a language-like model that can predict cognitive reactions in human participants better than an MRI. This can be used to continue study in behavioral science and neuroscience.
2025.01.03·arxiv.org·45 min read·TLDR AI This study suggests a method for reconstructing the genetic operation and designing a development framework for efficient parallel acceleration.
2025.01.03·arxiv.org·14 min read·TLDR AI This code uses advances in general 3D vision techniques to improve the niche of robot vision. Specifically, it can predict the dynamics of objects being manipulated by a robotic arm which then in turn improves the system's overall manipulation performance.
2025.01.03·github.com·GitHub repository·TLDR AI There are very few promising open replications of o1 reasoning. This work is quite promising in that it uses implicit rewards that don't require a formal reward method or model, but also rewards the outcome which is inline with what we know about reasoning models. The code is in progress, but its developers have released the data and models.
2025.01.03·github.com·GitHub repository·TLDR AI This repository contains a 3D object detection framework designed for single modality inputs, particularly catering to simplified and efficient scenarios.
2025.01.03·github.com·GitHub repository·TLDR AI A conditional diffusion model integrating visual and textual inputs designed to generate high-quality images conditioned on diverse text-visual contexts.
2025.01.03·github.com·GitHub repository·TLDR AI