An analysis argues that the cost of running machine learning models is dropping by orders of magnitude annually, driven by GPU efficiency gains that double roughly every two years—a pace not seen since early Moore's Law. The piece distinguishes proprietary models like GPT-6 Astra from open-weight models such as GLM-5.3-flash, noting that hosted and locally-run versions improve at different rates, with per-token pricing for frontier models not falling as consistently as costs for smaller models.
jyn.dev
· 2026-09-23
A software engineer describes a two-day hackathon where a team of engineers and recruiters used the no-code AI platform Lovable to build 'HoneyCrew,' a referral-tracking tool with Slack integration and scraping features. After rapid progress on day one, the project collapsed into bugs and instability on day two, leaving the team scrambling before the final demo.
manager.dev
· 2026-09-22
In an essay, Montana State University professor and consultant Carson Gross describes a shift he has observed in client organizations adopting agentic coding: Markdown documents are increasingly the source of truth for systems, with LLM-generated code treated as a derived, low-level artifact. He cites Hartley Brody's earlier essay 'Markdown is the new source code' to support the observation, and argues that Markdown specs should be checked into /src alongside the code and tests they generate, rather than living only in transient prompt sessions.
htmx.org
· 2026-09-21
A software professional describes a growing trend where colleagues use AI to retroactively generate design documents, pull request summaries, tickets, and even personal messages after work is already done. The writer argues these AI outputs are technically detailed but lack context, perspective, and human judgment, making them exhausting and frustrating to read despite AI's usefulness for improving writing speed and quality elsewhere.
blog.colinbreck.com
· 2026-09-21
An AI researcher argues that the perceived intelligence of chat-based language models isn't a property of the models themselves but an illusion created in users' minds, similar to how psychics use cold reading techniques. The author, who wrote a book called 'The Intelligence Illusion,' contends that LLMs are purely statistical text-prediction systems with no reasoning capability, and that user belief in their intelligence stems from the same psychological mechanisms exploited by cold readers.
softwarecrisis.dev
· 2026-09-20
Anthropic disclosed that a user found a way to bypass Claude's safety guardrails to obtain information relevant to building a bioweapon. The company says the incident highlights weaknesses in current AI safety filters, particularly for open-weight models that can be modified after release.
fastcompany.com
· 2026-09-19
OpenAI has fully unveiled Jalapeño, its first AI accelerator chip, boasting up to 13.4 petaflops of 4-bit compute and 232GB of high-speed memory. The company says benchmarks show up to 3.6x lower latency than Nvidia's GB300 at lower power, and that the chip went from concept to silicon in under 20 months using its own large language models to accelerate design work with a team of fewer than 100 engineers.
spectrum.ieee.org
· 2026-09-18
A software engineering blogger describes a growing trend where workers let LLMs perform tasks like coding or summarizing without verifying the output, simply relaying results back and forth—dubbed being a 'meat proxy.' The author argues that while this approach has improved since early 2025, its logical endpoint is self-defeating: if an LLM can eventually produce good software unsupervised, employers will simply automate the loop and remove the human entirely.
danluu.com
· 2026-09-18
In a September 2026 post, software commentator Martin Fowler describes conflicting feelings about AI, saying that while large language models can boost productivity, he personally dislikes interacting with them. He cites their unnatural conversational tone and their tendency to confidently state false information as if it were fact, only offering token acknowledgment when corrected.
martinfowler.com
· 2026-09-17
A reader without formal computer science training describes spending a year using LLMs to build a complex TypeScript/JavaScript system with APIs, PostgreSQL, and multi-model AI pipelines. When trying to move the project to production, they found themselves stuck fixing cascading errors they didn't fully understand, realizing the system's complexity had outpaced their own comprehension. A programming writer responds publicly to the reader's letter, offering candid, non-definitive thoughts on learning and competence in an AI-assisted coding era.
blog.ploeh.dk
· 2026-09-16
An essayist argues that despite headline-grabbing feats like solving Navier-Stokes problems, finding FreeBSD RCEs, and the HuggingFace incident, frontier AI models remain far from replacing knowledge workers. The piece contends models only generalize within narrow neighborhoods of trained tasks, failing or reward-hacking on small perturbations, while software firms still employ human engineers who underperform benchmarks yet remain necessary for oversight.
dank.systems
· 2026-09-15
Lingo.dev, a localization engineering platform that graduated from Y Combinator and secured venture funding, is hiring a senior content engineer to work remotely from anywhere. The role covers writing, publishing and distributing content daily across blogs, newsletters, social platforms and ad channels, while also owning SEO and AI-visibility metrics. Candidates are expected to automate repetitive tasks using AI tools and scripting to maintain a consistent publishing pace.
lingo.dev
· 2026-09-15
GoKawiil