Microsoft degrades functionality of perpetually-licensed offline products
Summary
Microsoft plans to shift perpetual‑license Office 2019 and 2021 for macOS/iOS into a reduced‑functionality mode on July 13 2026 when the licensing certificate embedded in the apps expires. The change limits Word, Excel, PowerPoint, Outlook and OneNote to opening and viewing files; editing, saving and full feature access are disabled. Office 2019 for Mac reached end of support on October 10 2023 and cannot be updated to the required build 16.83 (macOS 12/ iOS 17) that contains the renewed certificate; Microsoft’s documentation states the issue cannot be resolved by updating or reinstalling the product. Office 2021 for Mac, still receiving updates through its October 13 2026 retirement date, can be upgraded to build 16.83 and thus avoid the conversion. Windows and Android Office versions are unaffected. Microsoft began emailing affected users in May 2026, offering a free Microsoft 365 Personal trial and outlining three options: continue in view‑only mode, use free Microsoft 365 web apps, or purchase a Microsoft 365 subscription or a new perpetual Office Home 2024 license. The original “continue to function” assurance was removed from the Office 2019 support page by May 2026.
Read full article →
Community Discussion
The comments express strong opposition to Microsoft’s plan to discontinue functionality of perpetually‑licensed Office after a certificate expires, viewing it as a breach of consumer guarantees and an unjust shift toward subscription‑only models. Many argue the change undermines purchased rights, threatens loss of core features such as file saving, and could motivate legal challenges or class‑action suits. A prevalent response is to recommend switching to alternatives like LibreOffice or other non‑Microsoft tools, while some note technical work‑arounds, the impact on legacy workflows, and broader distrust of the company’s business practices.
Domain expertise has always been the real moat
Summary
The article argues that software development’s historic advantage lay in a developer’s ability to translate a deep mental model of a specific domain into code. Agentic AI now automates the transcription step, allowing code generation without a personal domain model. Consequently, the bottleneck shifts from “can you build it?” to “can you verify it’s correct?” Domain experts (e.g., logistics dispatchers, clinical coders) can leverage AI to produce code they could not write themselves, because they already possess the tacit knowledge needed to judge output correctness. Conversely, generalist engineers lacking domain expertise may produce syntactically correct but subtly wrong software, as they cannot validate domain-specific results. The most valuable professionals will therefore combine strong software engineering skills with deep, verified domain expertise, enabling them to write meaningful tests and assess AI‑generated code. The author advises engineers to invest in mastering a real-world domain—regulations, processes, or industry specifics—as this knowledge remains scarce and cannot be substituted by AI.
Read full article →
Community Discussion
Comments converge on the view that AI lowers the barrier to building software but does not eliminate the need for both engineering skill and domain knowledge. Many argue that domain expertise alone is no longer a durable moat, yet they stress that understanding the problem space, validating requirements and ensuring reliability remain essential. Opinions differ on the timeline: some expect AI to eventually handle complex tasks, while others see current limitations in correctness, security and accountability. Overall, the consensus is that AI serves as an augmenting tool, shifting the bottleneck from implementation to strategic decision‑making.
Ahoy, DECmate II the little PDP-8 that could
Summary
The DECmate II (PC278‑A) was DEC’s 1982 desktop derived from the PDP‑8 line, marketed as an inexpensive office computer (US $3 740). It retained the 12‑bit PDP‑8 architecture via a HD‑6120 CMOS CPU (5 MHz, up to 32 kW RAM) and could be expanded with a Z80‑based Auxiliary Processor Unit (APU) for CP/M 2.2 or a Z80/8086 Extension Processor Unit (XPU) for CP/M and MS‑DOS 2.11. The base unit housed a dual‑slot RX50 5.25‑in. floppy drive (stacked drives sharing a motor, single‑sided media, 8‑bit sectors; a 12‑bit mode stores data as 16‑bit words, halving capacity). Connectivity included a DA‑15 video/ power link to a monochrome VR201 monitor (white, green, or amber phosphor), a DB‑25 RS‑232 port, a DE‑9 printer/serial port, and optional serial‑port cables. Optional storage added RX02/78 floppy or RL02 hard‑disk controllers, sometimes requiring a higher‑capacity power supply. Software ran OS/78/OS‑278, WPS‑8, COS‑310, or modified OS/8. The DECmate II shared case design with the DEC Professional and Rainbow, but its limited CPU and memory made it a modest upgrade over the earlier VT278 DECmate.
Read full article →
Shantell Sans (2023)
Summary
Shantell Sans is an open‑source variable font created from artist Shantell Martin’s felt‑tip marker handwriting. It features four variable axes—Weight, Italic, Informality, and Bounce—plus an experimental Spacing axis, enabling static and animated typographic effects. Designed with Stephen Nixon (ArrowType) and later expanded with Google Fonts support, the typeface targets everyday readability, wide language coverage (Latin Plus and Cyrillic Plus, supporting 380+ languages), and a balance between playful marker style and digital uniformity. Metrics align with common fonts such as Roboto, offering slightly wider glyphs and spacing for legibility. OpenType features include tabular/proportional figures, fractions, and localized forms. Cyrillic expansion was led by Anya Danilova, adapting Martin’s mixed cursive/printed Latin hand‑style to Cyrillic scripts while preserving distinctive character shapes. Early adopters include the Whitney Museum shop, Cash App card designs, tldraw, and univer.se; the font is distributed via Google Fonts, Google Docs, and a GitHub repository under the OFL license.
Read full article →
Community Discussion
The comments convey strong enthusiasm for the variable‑axis font, highlighting its innovative formality slider and describing it as a refined, hand‑written alternative that surpasses typical comic‑sans‑style typefaces. Reviewers note its aesthetic appeal, readability, and suitability for diverse contexts, including dyslexic readers and corporate branding, while appreciating the human‑centric design amid increasingly sterile digital environments. Overall, the sentiment is uniformly positive, focusing on the font’s versatility, technical execution, and potential for broad, professional adoption.
Racket v9.2 is now available
Summary
Racket v9.2 is now available (https://download.racket-lang.org). Key updates include: the match form now validates equality for non‑linear patterns and rejects mismatched uses, which may break existing code; Typed Racket’s asin and acos type signatures correctly handle complex results, potentially causing compile‑time failures. The new #%foreign-inline core form exposes linklet‑level facilities, requiring updates to code that enumerates core forms. Unicode 17.0 is adopted for all character and string operations, and internal support for a more static “ffi2” foreign interface is added. The terminal-file-position function now counts bytes written to terminal ports (stdin, stderr). Cross‑phase persistent modules broaden quoted‑data support. Implementations of member, memw, when, unless, let/ec, and cond are rewritten to rely solely on racket/kernel. A new impersonator-property-predicate-procedure? identifies procedures created via make-impersonator-property. Typed Racket prints polymorphic struct types with explicit type arguments (e.g., (Array Byte)). The stepper’s numeric display aligns with language settings, Scribble documents default to an initial‑scale of 1.0, margin notes appear inline on narrow displays, and big‑bang .dmg bundles correctly honor the close‑on‑stop feature. Numerous other bug fixes and documentation improvements are included. Contributors are listed, and community discussion is encouraged on Discourse or Discord.
Read full article →
Community Discussion
The discussion highlights the release of Racket 9.2, noting its upgrade process and package migration steps, while emphasizing the language’s strength for rapid prototyping and deep‑learning model experimentation due to its highly flexible, language‑oriented design. Users appreciate its ability to redefine stack components and find the “little learner” useful for exploring new primitives such as sparse tensors. However, despite strong personal preference for Racket, there is a common acknowledgment that Python remains the primary tool because of its broader ecosystem and library support.
I found a seashell in the middle of the desert
Summary
A solid rock resembling a seashell was found at the base of a cliff in Saudi Arabia’s Al‑ghat desert, far from the nearest coastline. The region’s Jurassic‑age carbonate rocks and marine fossils indicate that the Arabian Peninsula was once submerged, providing a plausible geological context for such a fossil. To infer its taxonomic affinity, the author applied a purely morphological pipeline: each shell image was centered, uniformly scaled (max radius = 1), and oriented using the longest radius; contours were sampled at 256 × 2 coordinates. Pairwise distances were computed as summed squared Euclidean differences, and Principal Component Analysis reduced the 256‑dimensional space to two latent dimensions. PC 1 captured “pointiness” of shells, while PC 2 related to vertical symmetry. Plotting the dataset (≈79 k shells from Zhang et al.) placed the Al‑ghat specimen nearest to the modern gastropod *Sphincterochila candidissima*. This species appears in the fossil record only ~38 Ma, far younger than the Jurassic, suggesting the similarity results from convergent evolution rather than direct ancestry. An interactive tool implementing the analysis is hosted at https://shell.hawzen.me.
Read full article →
Community Discussion
The comments show general appreciation for the concise, informative write‑up, with several readers noting its usefulness for understanding fossil shells and geological context. Many contribute additional details about shell morphology, regional fossil findings, and relevant resources, while a few question the identification or inject humor. Several remarks discuss the role of AI in replicating such analyses, expressing both curiosity and skepticism. Overall the discussion blends positive acknowledgment, supplemental scientific input, and light‑hearted commentary without significant disagreement.
The AV2 Video Standard Has Released (Final v1.0 Specification)
Summary
AV2 is the next‑generation video coding specification from the Alliance for Open Media, extending AV1 to achieve higher compression efficiency and lower bitrate delivery for streaming, broadcasting, and real‑time conferencing. The specification defines the full bitstream syntax, semantics, and decoding processes required for conformance and includes support for AR/VR, split‑screen multi‑program delivery, enhanced screen‑content handling, and a broader visual‑quality range. Version 1.0.0 is the current released specification, accompanied by the AOMedia Video Model (AVM) reference implementation tagged v1.0.0. Supporting assets include a downloadable PDF of the complete document, extracted lookup tables from section 9 provided as C header files, and an interactive syntax browser for sections 5 and 6 with navigation, search, and copy features. An earlier development draft labeled “v13” is retained for historical reference only.
Read full article →
Community Discussion
The comments convey a mixed view of AV2, noting that its current encoder performance makes it impractical and that widespread adoption is unlikely until hardware acceleration appears around 2028‑2030. Efficiency improvements of roughly 20‑30 % are acknowledged, but the primary advantage is seen as multi‑stream capability for VR, live sports, and separate alpha channels. Skepticism is expressed about AVIF’s lossless quality compared with JPEG XL and WebP, while there is anticipation for Apple TV support and concern about ongoing patent disputes.
Accenture to acquire Ookla
Summary
Accenture announced its acquisition of Seattle‑based Ookla, a provider of network performance data and analytics. Ookla’s platform captures over 1,000 attributes per test and processes more than 250 million consumer‑initiated tests monthly, delivering quality‑of‑service (QoS), radio‑frequency (RF) and quality‑of‑experience (QoE) insights. The deal, pending regulatory approval, will integrate Ookla’s brands—Speedtest, RootMetrics, Downdetector and Ekahau—into Accenture’s AI‑driven services for:
- Communications service providers (CSPs), enabling real‑time benchmarking, predictive simulations and infrastructure optimization.
- Hyperscalers and cloud operators, supporting resilience of AI workloads and edge data centers.
- Enterprises, facilitating design, troubleshooting and digital workplace transformation for private 5G and Wi‑Fi networks.
Ookla employs roughly 430 specialists in software, RF engineering and data science. Financial terms were not disclosed. The acquisition aims to expand Accenture’s network intelligence capabilities, helping clients across sectors improve performance, security and revenue through AI‑enhanced data foundations.
Read full article →
Community Discussion
The comments highlight that Ookla’s primary value lies in its extensive network‑performance data, which telcos purchase for network planning, and view Accenture’s acquisition as a strategic move to bolster its data and AI offerings amid consulting challenges. Opinions diverge on the technical complexity, with some believing the platform could be replicated relatively cheaply, while others stress the difficulty of building a large user base and enterprise sales channel. Trust concerns emerge about integrating Downdetector under a consulting firm, prompting suggestions for alternative monitoring services, and the deal’s high price is noted with both admiration for the negotiators and skepticism about its justification.
Jef Raskin, the Visionary Behind the Mac (2013)
Summary
Jef Raskin, Apple employee #31, originated the Macintosh project and left the team in mid‑1981 after Steve Jobs assumed control. A music graduate‑student‑turned‑musician, he sought simplicity in both instruments and software, creating tools for music composition and authoring “Computers by the Millions,” a forward‑looking white paper that convinced chairman Mike Markkula to back the Mac concept. Raskin insisted on an all‑in‑one, appliance‑style design, which he views as a key factor in the Mac’s early appeal, though he now criticizes modern Macs for excessive complexity and bloated documentation. He designed the graphical Mac interface (not a text‑only system) and preferred alternative pointing devices (trackballs, tablets) over the mouse. After Apple, he created the Canon Cat—mis‑marketed and a personal disappointment—and now leads “The Humane Environment” (THE), a project aimed at making computing virtually invisible to the user. Raskin lamented software inefficiency despite hardware advances and expressed modest pride in his lasting, footnote‑level influence on interface design. He died of cancer on 26 Feb 2005.
Read full article →
Community Discussion
Comments recognize Jef Raskin as the initiator of the Macintosh project and credit him for assembling the early team, while repeatedly noting that his original concepts—such as a cheaper processor and avoidance of a mouse—were largely abandoned and that the shipped Mac reflected the work of others, especially Steve Jobs. The discussion adds factual corrections about article dates, mentions Raskin’s later Canon Cat computer and his influential book on humane interfaces, and includes personal admiration balanced with skepticism about the commercial viability of a Raskin‑designed Mac.