OpenAI published its research paper on GPT-2 in February 2019 and released the model in stages across the year: the 117M parameter version in February, the 345M version in May, the 774M version in August, and the full 1.558 billion parameter model in November 2019. The staged release was explicitly justified by OpenAI’s concerns about potential misuse for generating disinformation and spam at scale — the first time a major AI lab publicly withheld model weights for safety rather than commercial reasons. Read more
The daily life of a software engineer is usually less dramatic than movies make programming look. There are no montages of cascading green characters or last-minute keystrokes that save cities. Most days are a mixture of reading, testing, discussing, debugging, and — somewhere in the middle — writing code. Read more
Software engineers spend a lot of time learning from screens — documentation, source code, issue trackers, tutorials, and online discussions. Books offer a different pace. They allow one author to develop an idea fully, without competing with notifications and search results, and without the pressure to be immediately practical that tutorials often carry. Read more
Facebook published the Libra whitepaper on June 18, 2019, through its Calibra subsidiary, proposing a global digital currency backed by a reserve basket of low-volatility assets — approximately 70% USD, 26% EUR, and smaller allocations to JPY and GBP government securities and money market funds. The Libra Association, incorporated in Geneva, Switzerland, launched with 28 founding members (later expanded to 100 targeted) including Visa, Mastercard, PayPal, Stripe, Uber, Lyft, and Spotify, each contributing $10 million. Facebook’s own subsidiary Calibra (later Novi) would build the consumer wallet, while the Association would govern the protocol. Read more
If I were building a gaming PC on July 1, 2019, this is the machine I would actually want. I am not simply choosing the most expensive part in every category. The priority is gaming performance, then reliability, cooling and enough headroom to keep the system useful for years. Read more
The Raspberry Pi 4 Model B was released on June 24, 2019, in three RAM variants: 1GB ($35), 2GB ($45), and 4GB ($55) LPDDR4 — replacing the Raspberry Pi 3 B+ (launched March 2018) as the current flagship. The BCM2711 SoC used a quad-core ARM Cortex-A72 at 1.5 GHz, an out-of-order execution microarchitecture significantly more capable than the Pi 3’s in-order Cortex-A53 at 1.4 GHz. Sysbench CPU benchmarks showed roughly 2–3× improvement in integer throughput; the out-of-order pipeline allowed the Cortex-A72 to extract instruction-level parallelism that the A53 executed strictly in sequence. Read more
Google announced the Pixel 3a and Pixel 3a XL at Google I/O on May 7, 2019. The Pixel 3a launched at $399 (5.6-inch) and the 3a XL at $479 (6-inch) — less than half the Pixel 3’s $799 starting price. The phones used a Snapdragon 670 (Pixel 3a) rather than the Pixel 3’s Snapdragon 845, with a plastic back instead of glass, no wireless charging, and no front stereo speakers. They retained the same 12.2MP rear camera sensor and f/1.8 lens as the Pixel 3, the same pixel binning algorithm, and Google’s full HDR+ and Night Sight computational photography pipeline running in software on the Snapdragon 670’s DSP. Read more
Low-power wireless devices face a fundamental tension. They need reliable communication, but their batteries may need to last months or years in places where replacing them is difficult or expensive. Keeping a radio active continuously is not an option. 6TiSCH is a protocol architecture designed to resolve this tension — allowing constrained devices to participate in Internet-style networks while managing energy carefully. Read more
Microsoft announced the new Chromium-based Edge on December 6, 2018, and released the first Canary and Dev channel builds on January 20, 2019, followed by the Beta channel on April 8, 2019. The decision meant abandoning EdgeHTML — the proprietary rendering engine built from the remnants of Trident (Internet Explorer’s engine) and introduced with Windows 10 in 2015 — along with the Chakra JavaScript engine. At the time of the announcement, Edge held approximately 4% global browser market share versus Chrome’s ~63%, Internet Explorer’s ~7%, and Firefox’s ~10% (StatCounter data, Q4 2018). The compatibility situation was the primary driver: websites routinely displayed “Works best in Chrome” banners when accessed from Edge, and developer tools in Chromium’s DevTools were far more capable than Edge’s F12 tools. The extension ecosystem was another gap — the Chrome Web Store had tens of thousands of extensions while Edge’s own extension library had a few hundred. Read more
Google announced Stadia at GDC in March 2019, promising a cloud gaming service that would stream games running on data-center hardware directly to Chrome browsers, Chromecast Ultras, and Pixel phones. The platform launched November 19, 2019 in 14 countries with a Founder’s Edition bundle ($129) containing a Stadia Controller, a Chromecast Ultra, and three months of Stadia Pro ($9.99/month), which offered 4K/60fps streaming and a rotating selection of free games. A free tier at 1080p/60fps was announced but not released until April 2020. Read more
Google shut down the consumer version of Google+ on April 2, 2019. The shutdown had been announced October 8, 2018, when Google disclosed that a software bug introduced in March 2015 and patched in March 2018 had exposed the private profile fields of up to 500,000 Google+ accounts to third-party app developers through the People API — specifically, apps that were granted access to public profile data could also read private profile fields (name, email address, occupation, gender, age) that users had intended to share only with specific people. Google had not publicly disclosed this bug when it was discovered in March 2018 — citing that no evidence of misuse had been found — a decision criticized after The Wall Street Journal reported it in October 2018, suggesting Google’s non-disclosure was influenced by concerns about regulatory and reputational consequences following the Cambridge Analytica/Facebook scandal (March 2018). A second API bug was found in November 2018, affecting approximately 52.5 million users by exposing additional profile fields (name, email address, occupation, age) to apps that users had not explicitly authorized, accelerating the shutdown timeline from August 2019 to April 2, 2019. Read more
NVIDIA announced GeForce RTX 20-series laptop GPUs at CES in January 2019, bringing the Turing architecture’s dedicated RT cores for ray tracing and Tensor cores for DLSS into mobile systems. The mobile lineup included the RTX 2060, 2070, and 2080, with most high-end configurations also available in Max-Q variants designed for thinner chassis. Read more
Kubernetes 1.13 released on December 3, 2018, marking kubeadm — the official Kubernetes cluster bootstrapping tool — as generally available (GA) after nearly two years in alpha/beta. kubeadm init (initialize a control plane node) and kubeadm join (join a worker node to an existing cluster) provided a consistent, supported procedure for bringing up a production-grade Kubernetes cluster from bare OS without hiding the cluster behind a managed service abstraction. Prior to kubeadm GA, bootstrapping a cluster involved either running a vendor-managed installer (kops, Rancher, OpenShift) that made different decisions about network configuration and storage, or manually reading multi-page documentation to run etcd and the kube-apiserver, kube-scheduler, and kube-controller-manager as system services. kubeadm’s GA status meant distributions and tooling could declare a standard bootstrapping interface, enabling reproducible cluster setup across on-premises hardware, VMs, and bare metal. Read more
The Raspberry Pi Foundation released the Raspberry Pi 3 Model A+ in November 2018 at $25, ten dollars less than the Model B+. It brought the processor and wireless capabilities of the Pi 3B+ — the fastest model in the third generation — into the smaller Model A form factor, which had always prioritized size and cost over maximum connectivity. Read more
Machine learning is no longer something I hear about only in research papers. TensorFlow was open-sourced in November 2015. Keras made it more approachable — and Google officially adopted Keras as TensorFlow’s high-level API in 2017. PyTorch, released by Facebook’s research lab in September 2016, gained significant traction in academic research through 2017 and 2018 because its dynamic computation graph made debugging more natural. scikit-learn has made classical machine-learning algorithms available in Python with a consistent API since 2011. Cloud platforms — AWS SageMaker (launched November 2017), Google Cloud ML Engine, Azure Machine Learning — now offer managed environments for training and deploying models without running your own infrastructure. Read more
IBM announced on October 28, 2018 that it would acquire Red Hat for $34 billion in cash — $190 per share, representing a 63 percent premium over Red Hat’s closing stock price on October 26 — in the largest software acquisition in history at that point, surpassing Microsoft’s $26 billion LinkedIn acquisition in 2016. Red Hat had been founded in 1993 by Marc Ewing and Bob Young and had built its business on a commercial support and subscription model for open-source software: customers paid not for the software itself (which remained freely available) but for Red Hat’s enterprise support lifecycle, certification testing, and legal indemnification. By fiscal year 2018, Red Hat reported approximately $3 billion in annual revenue, growing around 20 percent per year. IBM CEO Ginni Rometty framed the acquisition as the cornerstone of IBM’s pivot to hybrid cloud — the model in which enterprise workloads run partly on on-premises servers, partly on private clouds, and partly on public clouds (AWS, Azure, GCP), with Red Hat’s technology providing a consistent operating environment across all three. Red Hat Enterprise Linux (RHEL) had become the dominant enterprise Linux distribution, running the majority of Fortune 500 production workloads. Equally important was OpenShift, Red Hat’s Kubernetes distribution, which Red Hat had built up through its acquisition of CoreOS (February 2018, ~$250 million) — giving IBM a leading Kubernetes management platform at a moment when container orchestration was becoming the standard for cloud-native application deployment. Read more
Apple announced the A12 Bionic on September 12, 2018 alongside the iPhone XS ($999), XS Max ($1099), and XR ($749 — releasing October 26). Manufactured on TSMC’s first-generation 7nm process (N7, not yet EUV-based) with approximately 6.9 billion transistors, the A12 was one of the first mass-production chips built at 7nm — Qualcomm’s Snapdragon 855 remained on 7nm TSMC but didn’t ship until December 2018, and Samsung’s 7nm processes weren’t available until 2019. The A12’s CPU used two Vortex high-performance cores (2.49 GHz) and four Tempest efficiency cores — Apple’s first chip with a heterogeneous big.LITTLE-style CPU arrangement, following ARM’s own design pattern — plus a four-core GPU that Apple claimed was 50% faster than the A11’s GPU. Read more
A computer looks complicated from the outside, but its basic operation can be described with a simple cycle repeated billions of times per second. Understanding that cycle helps explain why software behaves the way it does — why some operations are fast, others slow, and what actually happens when code runs. Read more
