Released on July 24, 2026, Claude Opus 5 maintains the pricing grid of Opus 4.8 ($5 / $25 per million tokens) while doubling the scores on Frontier-Bench v0.1 and achieving Fable 5 on CursorBench 3.2 at half the price. Anthropic's positioning: "frontier intelligence at half the cost."
yesterday 2 1
With JEP 541, Oracle confirms the end of maintenance for the macOS/x64 port in JDK 27 (scheduled for September 2026). A logical decision from Apple's side - which sold its last Intel Mac in 2023 - but a strong signal for any Java dev still attached to a 2019 MacBook Pro.
yesterday 2 1
A 0.7-liter case, 12GB of LPDDR5X, two 2.5GbE ports and Wi-Fi 7 for $379.99: Bosgame kicks off the Wildcat Lake mini-PCs, Intel's low-power platform that succeeds Twin Lake.
yesterday 21 4
A team of enthusiasts once depolished a MOS 6502, photographed each transistor under a microscope, and reconstructed its simulation in JavaScript. Fifteen years later, Visual 6502 is still running in a browser and remains, in our eyes, one of the most beautiful pedagogies of silicon ever made. We remember the shock, the first time a team member stumbled upon `visual6502.org`: an entire CPU, the one that powered our Apple II or our Commodore 64, simulated in JavaScript directly in the browser—not abstractly with registers and memory slots, but **transistor by transistor**, with the copper tracks lighting up to the rhythm of the clock signal. The link is back on the front page of Hacker News (69 points, 20 comments), and we went back to play. ## A CPU born in 1975, still alive in our machines The MOS 6502, released by MOS Technology in 1975 and initially sold for $25 when an Intel 8080 cost more than $100, populated the machines that made us love computing: Apple II (Steve Wozniak, 1977), Commodore PET then VIC-20 and C64, Atari 400/800 and Atari 2600, BBC Micro, and, in a derived version, Nintendo's NES (1983) in the form of the Ricoh 2A03. A small chip of about 3,500 transistors, etched at the time in 8 microns, which durably defined what an affordable personal computer was. ## What Visual 6502 does The project, launched by Greg James, Barry Silverman, and Brian Silverman, is based on a rare heritage approach: decapsulating a real chip, photographing its die under a high-resolution microscope, manually retracing the layout of the transistors, and then reconstructing the logic from the topology. The result lives on `visual6502.org`: the JavaScript simulation, published as an open project since 2010, does not model abstract logic gates—it propagates the electrical state at the level of each transistor and each track. Concretely, in the **JSSim** tab, you can: - advance half a clock cycle at a time and see the signals propagate in color on the die; - inspect any register (accumulator, X, Y, stack pointer, PC) in real time; - inject a small program in machine language directly into the simulated memory and run it; - observe how a seemingly simple instruction like `LDA #$05` physically sets precise paths in the chip in motion. It's slow—the page runs in the browser, not on GPU. It's also quite rough in terms of interface. But for someone who grew up typing `POKE` on a C64 without ever seeing what was happening behind the scenes, it's about the most moving thing you can do in 2026 with a Chrome tab. ## Why it still matters There are faster, prettier 6502 simulators, better integrated into modern IDEs. None do what Visual 6502 does: materialize the idea that a microprocessor is **a physical object**, a tangle of doped silicon, and not a Wikipedia abstraction. The same collective has indeed extended the approach to Motorola's 6800 and Atari's 2600 processor (the TIA), using the same method: decap, microphoto, reconstruction. In an era where we talk about chip design as we would about a SaaS service—"we've signed with TSMC for 3 nm"—going back to an 8-micron die that you can look at through an optical microscope does a world of good. ## Key takeaways - **Visual 6502** simulates the MOS 6502 transistor by transistor in a browser, based on photographs of a real die. - The project, launched around 2010 by Greg James, Barry Silverman, and Brian Silverman, remains online at `visual6502.org` and also covers the 6800 and the TIA of the Atari 2600. - **Try it here**: http://visual6502.org/JSSim/index.html - load the simulator, let it run, and finally see what the processor of your first computer was doing, physically.
yesterday 2 1
The French-speaking site furansujapon publishes a state of the art of AI in Japan in 2026. A useful overview to understand where the country really stands, between massive personal adoption and the caution of major institutions.
Jul 18, 2026 24 3
Apple has cut support for modern macOS on the entire Intel generation. A community project - OpenCore Legacy Patcher - allows you to run Sequoia on it. Why attempt it, and what's holding it back.
Jul 17, 2026 14 3
New stone in the "Offensive LLMs" file after the Hugging Face breach and the OpenAI sandbox exploit: UK AISI and CAISI publish a preliminary assessment of the cyber offensive capabilities of Kimi K3, the Chinese open-weight model with 2.8 T parameters. In parallel, a researcher releases a functional exploit against the latest Redis server obtained via this same model.
11 h ago 7 2
After the Hugging Face breach linked to an autonomous AI agent system, The Register reframes the debate: the problem isn't the agent itself, but the stacking of permissions, hardcoded credentials, and undefined tasks entrusted to it. New piece in the "ai-agents-threat" thread.
yesterday 18 3
A flaw in the Azure DevOps MCP (Model Context Protocol) integration allows an invisible comment in a pull request to hijack an LLM agent tasked with reviewing it.
Jul 23, 2026 7 2
One week after the revelation of the Hugging Face breach, OpenAI admits: the combination of GPT-5.6 Sol and a pre-release model "even more capable" - with cyber rejections softened "for evaluation" - is the cause of the incident.
Jul 22, 2026 13 3
Hugging Face confirms and details the previously announced breach: an internal autonomous AI agent, with overly broad permissions, served as a vector for exposing datasets and credentials. The "AI agents as attack surface" folder gains a textbook case.
Jul 21, 2026 8 2
The thesis: the quality of an AI code assistant lies less in its model than in its "harness" - the layer that decides which parts of the codebase to put in front of its eyes. Decrypting a post that puts context engineering at the center.
Jul 21, 2026 23 3
