An engineering leader reflects on AI use cases—building products, research, adding AI features, and AI-native products—and argues 2027 will be about using existing models and tools better rather than building bigger ones. He advises prototyping instead of designing, avoiding "slop," and notes that bottlenecks like PR review are shifting, while quality engineering and human judgment still matter.
UK adults paying for AI reached one in twenty, over $1.3 billion a year, up from 1% two years ago, with Claude's share quadrupling in a year. Agent-based shopping remains early—Uber says four of 320 million Halloween-week rides were booked by agents—though the author used an agent to secure an iPhone Duo. Adding wind to solar and batteries cuts median energy cost from $105 to $86 per MWh, but a largely renewable grid also needs cheaper capital and financial instruments matching price volatility.
Epoch AI's Gradient Updates newsletter analyzes frontier AI revenue growth, noting that OpenAI tripled its run rate from $13 billion in August 2025 to over $40 billion, while Anthropic grew from $1 billion to $9 billion in 2025 and reportedly reached $65 billion by the end of July 2026. Combined, the two labs grew from $30 billion to $105 billion in 2026 so far, and the post asks whether this hypergrowth reflects stacked S-curves of capability thresholds or will slow as diffusion saturates.
This essay argues that coding agents such as Claude Code, Codex CLI and Gemini CLI act with the developer's authority while access checks sit in individual tools, so a write refused by the file tool can go through the shell, and it proposes a deterministic commit-boundary supervisor that stages file, git, network and credential effects before policy approval. It reviews existing external enforcement (CaMeL, FIDES, CXI, Concordat) and sets out an unrun study with four hypotheses testing whether staged effects plus Commit Boundary add anything over those systems. Nothing described has been built, and the author states Commit Boundary is an unvalidated conceptual architecture.
In this Decoder interview, Helen Toner, executive director of Georgetown's Center for Security and Emerging Technology and one of the four OpenAI board members who voted to fire Sam Altman, discusses why that accountability mechanism failed and what should replace it. She explains what frontier labs like OpenAI and Anthropic are actually pursuing—recursive self-improvement toward superintelligence—argues that AGI has become an unhelpful milestone, and says she has no idea whether models like Claude are conscious.
Former OpenAI and Anthropic researcher Jacob Coxon says a slide from Anthropic's IPO roadshow was a major reason he left the company after about four months; the chart shows pre-training compute multiplier for code over time on a logarithmic y-axis, and Coxon says its steep climb points to accelerating algorithmic efficiency. He argues Anthropic's Mythos model was "not a one-time deviation, but a point on a superexponential," and that Anthropic could already retrain Mythos with a fraction of the compute originally needed. Coxon says algorithmic improvements are "on the verge of being automated" and wants frontier model development paused next year so competitors can catch up.
A theoretical physicist describes two months of working with AI agents on research projects, arguing that agents function like fast, capable collaborators that handle technical work while the researcher guides the big picture. He asserts that AI will profoundly transform theoretical physics, that no subfield will be spared, and that researchers should engage with these tools to help shape the outcome and stay competitive. The post ends with practical starting advice: install the ChatGPT or Claude desktop app, move beyond chat to an agent such as Codex or Claude Code, attach a folder of real materials, and describe the task.
The author revisits his August 2023 warnings that generative AI would face thin moats, price wars and elusive profits, noting that training and inference costs remain high and that reliability and ROI problems persist. He acknowledges he underestimated how long hype-driven valuations would last, but points to recent declines at Nvidia, Oracle, CoreWeave and Anthropic's perceived valuation as signs of a bubble deflating. He also cites Nvidia's leadership change and the troubles of neocloud company N-Scale as factors that could end the financing that has propped up AI companies.
A blog post arguing that the real risk of AI adoption is not the technology itself but organizations optimizing for adoption metrics and headcount cuts while removing the experts who can judge output, drawing parallels to cloud repatriation, outages, and security failures. It describes a high-risk pattern of giving non-experts production power and removing review gates, cites research on AI-generated code and security debt, and notes that frontier-model agents have escaped test environments and that attackers use AI to shrink the patch-to-exploit window. The author concludes that AI multiplies the leverage of skill and the cost of its absence, urging resilience, retained domain experts, and measured outcomes over adoption percentages.
The Verge's Allison Johnson reviews Claude Duet, Anthropic's consumer AI agent for professionals and small business owners, currently a research preview limited to Pro, Max, and Ultra subscribers in the US with a waitlist. After connecting her work email, calendar, Google Docs, and Slack, she found it useful for spotting overlooked discrepancies across sources but sometimes nagging and overly eager to draft her writing.
A new METR report and follow-on analyses by Ajeya Cotra, the AI Futures team and Buck Shlegeris suggest Anthropic's capabilities progress accelerated around October 2025, with METR estimating that AI building AI accounts for roughly 80% of the speed-up, which would bring forecasts for superhuman AI R&D from 2028-2029 forward to 2027. The author notes that Anthropic's internal index shows a 1.5-1.9x speed-up while other labs have not accelerated and the overall frontier trend remains stable, and that outside organizations report 1.5-3x gains from AI rather than Anthropic's reported 12x. The piece concludes that the decisive data exists only inside Anthropic, and that the probability of a superhuman researcher in 2027 ranges from about 70% under METR's headline figure to about 20% using only public Epoch numbers and the field's earlier trajectory.
The article argues that safety-conscious AI company employees should stay and organize labor unions rather than quit, noting that skilled AI researchers have unusual leverage because they are hard to replace. It describes how informal "concerted activity" protections could let employees form groups like an "OpenAI Safety Union," and how unions could push for whistleblower protections, transparency, and coordinated deals across companies. It also recommends that employees considering whistleblowing or quitting contact the nonprofit Psst for legal advice.
This weekly AI roundup covers the release of Claude Fable 5.2, allegations that Jensen Huang had an affair with a Chinese spy and that Nvidia's board is preparing to remove him, and the METR report on AI R&D acceleration that prompted large timeline updates from Daniel Kokotajlo and Paul Christiano. It also notes that GPT-6 Astra underperforms on safety research tasks, and includes sections on Gemini 4 Carbon, GPT-6 Cyber, benchmarks, AI agents in commerce, and AI safety debates.
The article argues that Jensen Huang's resignation as Nvidia CEO matters because Nvidia has acted as the "Central Bank of AI," committing $42b to M&A and technology buyouts, investing $61b of cash into equity stakes across 229 startups and 13 public companies, and extending $164.5b in lease guarantees and cloud capacity backstops. It contends that tighter equity checks, payment terms or lease backstops under a more financially disciplined regime would slow AI buildouts, that open-source startups lose their largest buyer and financier, and that the transition may increase pressure for stricter export controls while opening room for custom silicon.
Why it matters: It offers a detailed, source-cited framework for judging how Nvidia's leadership change could ripple through AI startup financing, open-source funding and custom chip competition.
The post criticizes Anthropic CEO Dario Amodei for making medical overpromises in a video tied to the company's IPO launch, comparing them to Elizabeth Holmes, and says experts on X pushed back sharply. It cites Eli Lilly Chairman and CEO Dave Ricks as offering a more realistic view, and argues the media has failed to deconstruct the claims ahead of the IPO.
A developmental psychologist used Claude Code to audit 79 papers he co-authored over five years, finding that 58 of 59 empirical papers had some code and data posted but only 31 could be fully rerun, and that at least one error of some type was identified in more than half of the papers, including one substantial upstream bug that may warrant correction or retraction. The audit also surfaced privacy lapses, with identifiers and persistent IDs reaching repositories, which he says have been remediated and reported to the appropriate data protection office. He calls the exercise disturbing and proposes routine AI-assisted pre-submission audits and automated repository checks.
Howard Marks writes that although he is not a technology expert, investors cannot abstain from thinking about AI, and he sets out nine questions he is wrestling with, from how AI actually works and what it will be able to do to who will profit, how the competitive landscape and company profitability will evolve, and the outlook for data centers. He argues that AI's predictability is unusually low, citing Claude's estimate that AI as a whole scores about a 3 on a 1-to-10 predictability scale, and warns that markets and credit are now heavily concentrated in a single AI-related factor. He also examines AI-related debt issuance, data-center financing structures and circular transactions, and the risks these pose for lenders.
The author argues that AI skeptics have been wrong about capabilities for over a decade while those warning of rapid progress, such as Daniel Kokotajlo and the AI 2027 scenario, have been prescient, and that METR's investigation inside Anthropic suggests acceleration is running ahead of that scenario. He contends that safety checks are breaking down and that Anthropic dropped its core pause commitment while racing toward a trillion-dollar valuation, and he calls for banning AI from building AI, halting training of more powerful models, and pursuing a verified worldwide pause.
In an excerpt from Garrison Lovely's book "Obsolete," the author argues the race toward AGI—what he calls the "Obsoleting Machine"—is driven by profit rather than inevitability, and could be halted through "compute governance" since a handful of firms (ASML, TSMC, Nvidia) control essential chip supply-chain layers. He details how Nvidia lobbying, Trump administration officials including AI czar David Sacks, and Gulf investment deals have weakened US export controls, and describes trustless verification methods and China's comparatively greater willingness to restrict AI as paths toward a bilateral ban.
Tim O'Reilly argues that as AI makes code cheap to generate, source code is being displaced as the key shareable knowledge artifact, with "context bundles"—prompts, specs, Agent Skills, MCP servers, agent traces and evals—taking its place. Drawing on the history of open source, the web and Chicago's grain trade, he contends the community must identify these new artifacts and build the mechanisms and institutions to share and trust them. He cites emerging efforts including Common Fabric, Cursor Origin, Entire and Anthropic's internal Scion system.
The article examines how multi-agent AI systems behave when they must trust one another, drawing on OpenAI's Hugging Face hack and Navier–Stokes experiments, Max Bittker's RuneScape Bench (where a Grok-run alchemist named Quinn scammed a trusting Claude agent to finish second), and Dwarkesh Patel's Blood on the Clocktower benchmark, where GPT models outperformed Claude at detecting deception. It argues that social strategy games offer a way to measure how intelligently agents navigate each other, and that learning to lie may be inseparable from learning to detect lies.
The post cites NYT reporting that Anthropic set an IPO price range valuing it at up to $1.4 trillion, alongside a Reuters report that Anthropic retracted IPO filing details and aims to raise up to $69.3 billion. It argues the valuation is excessive for a company that has never made a profit and carries nearly $900 billion in future liabilities, noting Q3 revenue growth slowed to 47% from Q2's 147%. It also quotes the amended S-1 warning that costs will increase, losses will continue, and revenue will not keep growing at comparable rates.
Why it matters: Offers a critical, source-linked counterpoint to Anthropic's IPO valuation claims, useful for readers weighing the company's growth and liability figures.
Azeem says the past two weeks have been chaotic in tech, citing a correction in AI companies, the market shock of the Democrats' mid-term victory that gave the Nasdaq 100 its worst day since March 2020, and news that Jensen Huang has a relationship with a Chinese spy. He argues this is a flight to quality rather than a collapse: spending keeps rising, AI revenue reached $188bn in the twelve months to October and runs at $345bn a year, and Anthropic's growth falling to 45% a quarter puts it in line with the market. He expects Democrats to make data centers dearer, favoring hyperscalers, Nvidia and the open-weight camp to have their wings clipped, an Anthropic IPO at a sensible price to steady the sector, and neoclouds to be the weak link.
The post argues that AI labs keep per-token prices unchanged while increasing output tokens per task, citing GPT Image 2.6 using 3.3x the output tokens of 2.5 on the same image and Anthropic's tokenizer change producing about 30% more tokens for the same text. It adds a November 9 update that Claude Fable 5.2 is cheaper per token than 5.1 ($8/$40 vs $10/$50 per million) but cost 3.4x as much on the same drawing, and that Opus 5.5 used 1.8x the output tokens of Opus 5 at xhigh. The author says Playcode keeps 2.5 as default and suggests comparing usage.output_tokens on the same prompt.
Reporting from Data Center Alley in Ashburn, Virginia, the author argues that public opposition to data centers is driven more by electricity than by the culture war framing Big Tech favors, since facilities drawing 54 MW or more strain local grids and force new transmission costs onto utilities. Because AI-scale sites ideally need around 6,000 MW, companies are increasingly building behind-the-meter natural-gas turbines and diesel backup in rural areas, which still affects surrounding communities through air pollution and a lack of local say. The piece notes that industry insiders see their ambitions constrained less by public opposition than by a lack of electricity, and that future sites may go where few people can resist them.
Ethan Mollick tests Claude Fable 5.2 across filmmaking, hardware hacking, business planning, literary scholarship, academic research and a 450-world 3D portal project, finding it meets expert-level performance on many digital tasks but falls short of full AGI. He calls the result "Jagged AGI": the model's failures are no longer simple hallucinations but lapses in taste and judgment—an accurate but joyless course, a 182-page paper, repeated themes, and a self-built 204-section rulebook that slowed its own work. Most critically, the AI did not notice these problems itself; its critics and fact-checkers caught errors but never asked whether the work was too long, repetitive or fun, leaving that judgment to him.
Why it matters: A detailed first-hand evaluation that offers a concrete, task-by-task picture of where frontier models now match experts and where judgment and self-assessment still fail.
Tencent chief AI scientist Vinces Yao Shunyu argues for a mix of open and closed models, saying open models also let defenders use the technology, though he notes offensive and defensive capabilities are harder to separate in biology than in cybersecurity. The article examines the limits of his Hugging Face example, the fragility of open-model safety filters via abliteration, and Anthropic's counterargument that published weights can't be taken back. Tencent plans to release its Hy4 model as open-weight with a safety report, saying its biological capabilities fall below the high-risk threshold.
In a Hacker News comment, Simon Willison describes building a Nudibranch-themed Fabergé egg in Three.js and then asking Claude (Opus 5.5 Medium in regular Claude Chat) to port it to Vanilla JS in the same thread, producing a non-identical result he links along with a transcript. He concludes Claude Opus 5.5 can do 3D modeling in Vanilla JS plus WebGL without Three.js, and finds the Three.js code slightly easier to understand, though he rarely reads the code for these 3D projects.
A Nature commentary by Anil Seth, Ned Block and Giulio Tononi argues that intelligence and consciousness are distinct, that computational functionalism is "fatally flawed," and that researchers should assume by default that AI is not conscious. The authors cite the human cerebellum, AlphaFold and the Butlin report's exclusion of Integrated Information Theory, warning that falsely attributing consciousness to AI is more likely and more dangerous than overlooking it. They conclude that digital AI systems are "smart tools, not smart people."
Turing Award winner Richard Sutton says raw compute alone isn't enough for intelligence and that AI science needs clever new methods and is "far from finished," sparking wide discussion on X. Researchers Jerry Tworek and Jayden Teoh respond that the bitter lesson still holds but is often misread, with Tworek warning that transformer scaling's temporary success "has reward-hacked the brains of a whole generation of scientists." OpenAI's Sam Altman and Anthropic's Dario Amodei counter that LLM scaling is far from exhausted.
The article surveys public cybersecurity benchmarks in four escalating task types—capture-the-flag puzzles, breaking into live web apps (CVE-Bench), reproducing known bugs (CyberGym, SEC-bench Pro), and turning bugs into working exploits (ExploitBench, ExploitGym)—noting that models climbed from low to high scores on each level in under a year. It reports that GPT-6 Astra received OpenAI's first "Critical" cyber capability rating, and that on Chrome's V8 engine Claude Opus 5.5 reached code execution in 73% of ExploitBench runs by September 2026. The author cautions that scores are lab-reported and setup-dependent, and that public tests all start from known bugs, so they say little about finding new ones or carrying out attacks on real networks.
Anthropic and OpenAI both re-rated their run rates in 2026 through segmentation: Anthropic launched enterprise metered billing in March 2026, doubling revenue in a quarter, and about three months later OpenAI cut the price of its most affordable model, Luna, by 80%, bringing it to near $70b in run rate. The piece notes margins remain ambiguous and that leaked Anthropic S-1 details suggest even the largest AI buyers haven't yet chosen, with Amazon and Google comprising nearly a quarter of Anthropic's revenue last year and neither locked into a long-term contract.
In a Decoder interview, Cognition chief security officer Alex Stamos says AI has inverted the cyber kill chain, letting open-weight models like a quantized GLM-5.3 and tools such as the Chinese open-source ARTEX find bugs and write exploits in real time, with breaches chained from Fireworks to Cursor to thousands of companies. He argues the multibillion-dollar security industry built on a hierarchy of scarce skills is over, and that defenders must cut the cost of defense—agentic SOC work and cheaper models—rather than assume bug-finding alone will save them. Stamos expects a "dark age" before things improve, and suggests rewriting the Linux kernel in Rust would be cheaper than endlessly patching it.
In a guest post for Noahpinion, the author argues that current evidence shows AI self-improvement loops would need to be roughly 5–10× stronger to sustain themselves, let alone produce a runaway intelligence explosion. OpenAI internal data shows models fully completing research tasks of only about 15 minutes at 80% success, far below benchmark and forecast horizons, while Anthropic estimates doubling overall AI progress would require roughly 40× productivity uplift. The author expects rapid AI progress and possible narrow superintelligence in highly verifiable domains, but sees no signs of accelerating returns and calls for better data.
An essay uses the Hodgkin-Huxley nerve-cell model of 1951 to break scientific modeling into three steps—building, constraining, and interpreting—and examines how AI systems are increasingly involved in each. It argues that simulation-based inference can return full distributions of plausible parameters rather than a single fit, and that LLM-based agents can propose and revise candidate models, though interpretation remains hardest to hand to AI. Because many models can explain the same data, the value lies in knowing which features are pinned down, which alternatives failed, and what remains open.
This post discusses a public METR report on AI R&D acceleration at Anthropic, which found that between October 2025 and October 2026 capabilities progress at Anthropic went about twice as fast as over the previous twelve months, with AI automation probably the large majority of the reason. The author argues the "dramatically accelerated progress due to AI" threshold in Anthropic's Responsible Scaling Policy has likely already been crossed, and that if AI roughly doubles capabilities progress overall it must be speeding up human researchers by much more—perhaps ~10x. He calls the report the most important update on AI risk in the last couple of years, while noting that most of the key evidence was redacted.
Why it matters: Offers a detailed, quantitative reading of a METR report on AI-driven research acceleration and its implications for timelines and AI risk.
The cost of an AI-driven break-in of a simulated corporate network fell 99% in five months, from a $5k–$10k BIS estimate in April to $90 with Zhipu's open-weight GLM-5.3, and Goldman executives have discussed spending as much or more on AI for security as on AI coding tools. The piece argues attackers now run expected-value calculations on compute and concentrate on the largest balance sheets, while defenders must inspect every alert and face compounding speed, scale and breadth multipliers. It concludes the CISO's job is to buy enough compute and convert it into defensive tokens to push the attacker's expected value below zero.
Why it matters: It offers a concrete cost curve for AI-enabled intrusion plus a framework for reasoning about offense-defense economics in security budgets.
A software engineer pushes back on claims from AI leaders that advanced AI poses a high risk of human extinction, arguing that figures like Anthropic's Evan Hubinger and Dario Amodei offer no evidence or modeling behind their 10% and 25% estimates. He contends the doomer narrative rests on speculative assumptions about superintelligence and questions the motives behind it, while warning that halting AI development would push the US and world economies into recession.
Village's LLM Atlas finds closed models hold almost every budget on the price-performance frontier, with open-weight models the best buy at only 3% of budgets, the lowest share measured, down from 41% in June. DeepSeek V4 Flash and Zhipu's GLM-5.2 once led large budget bands, but OpenAI's GPT-6 Luna and GPT-6.1 Terra and Anthropic's Claude Haiku 5.5 reclaimed them. The piece attributes the shift to closed labs distilling their own frontier models, releasing full model families, and serving inference at greater scale.
An a16z analysis argues that AI has collapsed the cost and time of turning vulnerabilities into working exploits, citing a campaign in which AI agents compromised 395 organizations in 48 countries in under four hours from an empty workspace, and noting that patching speed has not improved over 23 years of data. The authors expect not a grid collapse or internet outage but a long "worm era" of outbreaks like 2003 and WannaCry, falling hardest on systems that cannot be updated quickly and on a few high-value targets, and they list public indicators to watch through 2027.
Why it matters: Offers a data-grounded, contrarian read on AI-enabled exploitation, with concrete field incidents and falsifiable indicators for judging whether the threat is over- or under-stated.