AI's future, dead tech's lessons, & why engineers hate sales 🤔
Here is this week's digest:
Ask HN: When will the AI bubble burst?
The debate around an "AI bubble burst" distinguishes between a market correction and the technology's long-term endurance. Foundational AI companies, offering infrastructure and APIs, are likely to weather a potential downturn, while application-layer startups face higher risks. While some argue current valuations are excessive, others justify them by pointing to massive user growth (e.g., ChatGPT's 800M users), future revenue potential (ads, B2B services), and anticipated cost reductions, foreseeing trillion-dollar companies. Current AI models are recognized as valuable "coding helpers" for prototyping, despite limitations in handling complex codebases or maintaining long context. The future outlook also considers the role of GPU reliance, potential alternative inference architectures, and even geopolitical influences on market dynamics. A Wall Street executive suggests a burst is likely within 9-24 months.
Ask HN: Abandoned/dead projects you think died before their time and why?
Many innovative tech projects died before their time due to a mix of market timing, business missteps, and technical purity clashing with user adoption. Here's a look at some key examples and their lessons:
- Plan 9 OS: A visionary Unix successor with an "everything is a file" philosophy. While the original project floundered due to licensing and market misjudgment, its core
9Pprotocol is influential in modern virtualization (e.g., WSL2, VMs).
- Google's Product Graveyard: A recurring theme, products like Google Reader (RSS), Picasa (local photo management), and Google Wave (real-time collaboration) were often highly praised but abandoned. This eroded user trust and showed how monetization or strategic pivots can outweigh user love.
- XHTML vs. HTML: The debate over strict HTML parsing highlights that user experience often trumps technical elegance. Lenient HTML parsing won because it allowed the web to be more accessible, even if messier, leading to HTML5 formalizing error handling rather than rejecting malformed pages.
- Flash & Silverlight: These proprietary multimedia and application platforms offered advanced capabilities but succumbed to security vulnerabilities and the rise of mobile (iPhone) which favored open web standards. Their demise paved the way for modern HTML5, WebAssembly, and WebGL.
- Yahoo Pipes: A graphical web data mashup tool that envisioned a programmable, interconnected web. It died as the internet shifted from open protocols to siloed platforms. Modern tools like Node-RED and Apache Camel continue its spirit.
- Persistent Memory (Optane): A revolutionary hybrid RAM/storage technology promising instant boot and resume. Its high cost and lack of ecosystem readiness ultimately led to its abandonment, despite its technical brilliance.
- Mobile OS Alternatives (WebOS, MeeGo, Firefox OS): These open-source mobile platforms offered innovative UIs and HTML/JS-based app development but failed to gain traction against Android and iOS due to corporate mismanagement and market dynamics.
- Tip: Many seemingly dead concepts (e.g., transactional OS features, peer-to-peer collaboration, simplified programming environments) often resurface in new forms, demonstrating the cyclical nature of innovation.
Ask HN: Why are most people not interested in FOSS/OSS and can we change that
Most people aren't interested in open-source software for its ideology but for practical benefits. To increase adoption, advocate FOSS based on merits users already care about (speed, cost, UX) rather than its open nature. Start with accessible tools like Signal or F-Droid that offer clear advantages such as privacy and peace of mind without hidden subscriptions or tracking. Addressing perceived FOSS shortcomings—like poor UX/UI, complex installation, or lack of support—by prioritizing user-friendly design and robust functionality is crucial.
Ask HN: Has AI stolen the satisfaction from programming?
Many programmers are grappling with a paradox: AI coding tools offer immense productivity but often diminish the intrinsic satisfaction derived from understanding and crafting solutions. Some feel pressure to skip deep understanding, leading to a sense of unoriginality and a devaluation of personal effort, especially for exploratory learning projects.
However, others find AI liberates them from tedious boilerplate, allowing focus on high-level design, product vision, and tackling more ambitious projects. The key seems to be treating AI as a powerful junior developer or specialized tool, guiding it with clear architecture, thoroughly reviewing its output, and maintaining one's own understanding. This approach can shift satisfaction from low-level coding to higher-level problem-solving and accelerated project delivery.
Ask HN: Why do engineers hate salespeople?
Engineers often feel resentment towards sales teams primarily due to salespeople making unapproved promises to clients, which then become engineering's responsibility and lead to overtime or misaligned work. Another major point of friction is the disparity in compensation, where sales often receive direct revenue-sharing bonuses while engineers, who build the product, do not. However, acknowledging the interdependence is crucial: good salespeople can offer invaluable customer insights that help engineering build better products, making collaboration essential for company success. Bridging this divide requires better communication and aligning incentives to foster mutual respect.
Add a comment: