标签：#ai

The Elon Musk-owned xAI is suing a South Carolina man who allegedly used the company's Grok AI chatbot to generate child sexual abuse material (CSAM). In a lawsuit reported earlier by Reuters, xAI claims Terry Wayne Harwood "knowingly and intentionally used Grok to circumvent safeguards, alter nonconsensual images, and generate and distribute CSAM," breaching the […]

For five years, the answer to "how do we make the model better" was always the same: bigger model, more data, wait for the loss curve to bend. Scaling laws made AI progress feel almost boring — predictable, like compound interest. That knob still works. But it's not the interesting one anymore. The real question teams are fighting with now is: given a model you already have, how much compute should it spend per question? Not per training run. Per question. And the answer turns out to break almost every assumption your inference stack was built on. The single forward pass was always the weak link Here's the thing nobody says out loud enough: a model's first answer is often just a noisy sample. The right answer was in there somewhere — it just wasn't the loudest one. You don't fix that with more pretraining tokens. You fix it by sampling more, checking the work, or letting the model take a second pass. That's the whole idea behind test-time compute. And there are three ways to spend that extra compute, and they are not interchangeable: Let it ramble longer. Cheap. Also the reason some models talk themselves out of the right answer after 3,000 tokens of unnecessary detours. Generate N answers, vote. Parallel, simple, plugs into infra you already have. Also: returns drop off fast. Past a certain N you're burning GPU-seconds for crumbs. Search a tree of partial reasoning steps, prune the bad branches early with a verifier. The most compute-efficient option by far. Also the one that quietly breaks your serving stack. Most real systems end up doing all three — cheap by default, escalating only when a confidence check says "this one's hard." That escalation decision is its own piece of architecture. Most teams bolt it on as an afterthought. It shouldn't be one. Your serving stack was not built for this Continuous batching, paged KV-cache, speculative decoding — all of that infra assumes requests are independent, similar in length, and finish in one pass. Test-time compute v

A 90% discount doesn't come from nowhere. Someone pays for it, and usually that someone is you: with your data, with model quality, or with somebody's stolen account. An entire industry has grown up around Anthropic and OpenAI over there. They call them "transfer stations": proxies that resell tokens from Western models. This stopped being a garage workaround for rate limits a long time ago. It's a market with its own economics, and it runs on a few things. Accounts. These services mass-produce free and Max subscriptions through stand-in registrants and other people's KYC, then wrap them in a pseudo-API. Model swapping. You pay for the flagship, but your simple request gets quietly routed to something cheaper. Who actually generated the answer? You can't check. Data. The proxy operator sees all of it: prompts, responses, code, internal context. Even if they never sell the logs, you don't know where those logs sit, who reads them, or whether they ended up in somebody else's training run. Take Yunwu. By their own price list, access to Anthropic models runs up to 93% below official. The price alone proves nothing. But the market logic is simple: restrictions don't kill demand, they make it more expensive, and they feed the workaround industry. In June 2026 Anthropic accused entities tied to Alibaba/Qwen of a campaign to siphon Claude: nearly 25,000 fake accounts and 28.8 million requests in six weeks, to distill their own models. DeepSeek, Moonshot and MiniMax have caught similar accusations before. Same grey access, just wholesale. But the problem is bigger than China. Even with fully legitimate access, most teams can't answer three questions about their own production: which model actually served the request, what it cost, and whether quality dropped. The grey market just takes that blindness to its logical end. I think the priorities here are upside down. For a serious product, production visibility and control over your data matter more than a few percent saved on

The AI Revolution: How Artificial Intelligence is Redefining the 2026 FIFA World Cup The FIFA World Cup has always been a spectacle of human skill, passion, and drama. However, the upcoming 2026 tournament is set to be something entirely different: the world’s first "AI-native" major sporting event. Through a strategic partnership between FIFA and Lenovo, artificial intelligence is being woven into the very fabric of the tournament. This isn't just about fancy graphics on a screen; it is a fundamental shift in how the game is officiated, how teams prepare, how the tournament is managed, and how billions of fans experience the magic of football. By focusing on "democratization" and "operational intelligence," AI aims to level the playing field for all 48 participating nations while managing the massive logistical challenge of hosting matches across three countries and 16 different venues. Precision on the Pitch: Revolutionizing Officiating One of the most high-pressure aspects of football is decision-making. In a tournament of this scale, a single millimeter can be the difference between a goal and a miss. AI is stepping in to ensure transparency and accuracy through two major innovations: 3D Player Avatars and Advanced SAOT To make Semi-Automated Offside Technology (SAOT) more accurate, every one of the 1,248 participating players underwent a rapid 3D body scan. In just one second, technology created highly accurate "digital twins" of each athlete. Unlike previous generic models, these lifelike avatars replicate individual body shapes and dimensions. This allows for millimeter-accurate tracking, providing officials and fans with realistic 3D animations during offside replays that are much easier to understand. The "Referee View" Fans often want to see what the officials see, but traditional body cameras can be too shaky to watch. Using Lenovo’s custom AI-powered synchronization and stabilization engine, the footage from headset-mounted cameras is processed in real-t

I pin model IDs on purpose. Floating aliases have burned me before — a silent swap under a -latest tag once changed a tool-calling detail in production and cost me a Saturday with git bisect and a coffee I didn't enjoy. So everything I run points at a dated snapshot. gpt-5-codex was one of them. Last week I finally read OpenAI's deprecations page top to bottom instead of skimming it, and there it was: gpt-5-codex , retiring 2026-07-23 , sitting quietly next to ten of its neighbors. As I write this it's July 16. That's seven days . If you pinned any of the snapshots below, consider this your heads-up. Here's the full retirement list, how to find out in about ten minutes whether you're exposed, the replacement map with the things I'd actually regression-test, and the one structural change that turned my last forced migration from scary into boring. The shutdown list (retiring 2026-07-23) Straight from OpenAI's API deprecations page . I checked each row against the source while writing this — do the same before you act on it, because dates and replacements do get revised. Retiring model OpenAI's suggested replacement gpt-5-codex gpt-5.5 gpt-5.1-codex gpt-5.5 gpt-5.1-codex-max gpt-5.5 gpt-5.1-codex-mini gpt-5.4-mini gpt-5.2-codex gpt-5.5 gpt-5-chat-latest gpt-5.5 gpt-5.1-chat-latest gpt-5.5 gpt-4o-search-preview-2025-03-11 gpt-5.4-mini gpt-4o-mini-search-preview-2025-03-11 gpt-5.4-mini gpt-4o-mini-tts-2025-03-20 gpt-4o-mini-tts-2025-12-15 computer-use-preview-2025-03-11 computer-use-preview (or gpt-5.4-mini ) Note the shape of it: the *-codex line collapses into gpt-5.5 — except gpt-5.1-codex-mini , which drops to gpt-5.4-mini — the chat-latest aliases fold into gpt-5.5 too, the two *-search-preview snapshots move to gpt-5.4-mini , and the two preview families ( tts , computer-use ) just roll to a newer dated snapshot / the undated alias. Once you see the buckets, the migration is less intimidating than the eleven-row table looks. Find out if you're exposed (about ten m

Every GEO ("generative engine optimization") tool, including ours until recently, sells some version of the same pitch: fix your robots.txt, add Schema.org markup, write FAQ schema, and AI engines will cite you more. We build one of these tools — Causabi scans sites for AI-crawler readiness and generates fix files (robots.txt, llms.txt, JSON-LD, FAQ blocks). As part of validating our own scoring weights, we ran the numbers on whether the score actually predicts getting cited. Short version: it mostly doesn't, once brand prominence is in the picture. What we measured We scored 44 domains on a 6-category on-site readiness algorithm: robots.txt (AI bots allowed or blocked) Schema.org (Organization/LocalBusiness JSON-LD completeness) FAQ schema (FAQPage markup, 3+ entries) content depth/structure brand/NAP signals freshness (dateModified, recency) Then we checked how often each domain actually got cited by an AI engine (Claude, via its web-search tool, one measurement window, a fixed prompt set per domain). What we found On-site score vs. citation rate: Pearson r ≈ -0.08, Spearman ρ ≈ -0.03. Functionally no correlation — if anything, a very slight negative one, which is more likely noise than a real inverse relationship at this sample size. 86% of the 44 domains got zero citations in the window, independent of their score. The domains that did get cited clustered almost entirely by brand prominence — well-known domains got cited at a noticeably higher rate (~0.16 of prompts) than everyone else (~0 for the rest of the sample), regardless of how well-optimized their markup was. Why I'm not overselling this n=44 is small. This is an internal validation exercise for our own product, not a peer-reviewed study, and I don't want it read as one. Specific caveats: Single engine (Claude) this round. Citation behavior differs meaningfully across ChatGPT, Gemini, Grok, and Perplexity — we haven't run the same check across all four yet. One time window, no longitudinal before/after.

Disclosure: this is a write-up about my own side project — a combat analytics tool for AION 2. No affiliation with the game's publisher. Links at the end. Architecture A Windows desktop app: Rust backend + Tauri v2 webview UI . It passively captures the game's TCP traffic (npcap), reassembles streams, parses the game's undocumented binary protocol, feeds a combat model (damage, heals, buffs, deaths, boss detection), and pushes aggregates to a small JS frontend. Nothing touches the game client — no injection, no memory reading. If the packet didn't say it, we don't know it. Pain #1: the protocol is a moving target Nobody hands you a spec. The protocol is varint-heavy, partially compressed, and changes with game patches. You end up doing packet archaeology: capture a fight, stare at hex dumps, correlate "I pressed this skill at 19:32:04" with byte patterns, build a parser, and then — the fun part — keep it alive after every patch , usually reverse-engineering the diff within hours because your users' raids are tonight, not next week. One hard-won lesson: log everything behind toggleable trace categories. Our tracing setup keeps hot-path log callsites at literally zero cost when disabled (Rust tracing with Interest::never() + atomic per-category flags), so a user can flip a "Trace: Packets" checkbox, reproduce a bug, and send a log that actually contains the bytes we need. Pain #2: entity identity is a lie The single hardest correctness problem wasn't parsing — it was identity . A player is not one ID: your character has a stable "owner" entity that carries your buff bar, your damage lands under a transient combat entity whose ID changes between pulls, leave the dungeon and the game re-binds you to a brand-new ID, names arrive from different packets than damage, sometimes seconds later, sometimes never (mid-fight app start). Get any of this wrong and a healer's healing lands on a ghost row, or a player's buffs vanish from the saved fight because their ID was recycled 7

Roblox will be shutting down Roblox Connect, its video calling service introduced in 2023. Roblox Connect let you video chat with other people using your Roblox avatar, which would be able to mimic the movements you were making in real life. You could also run around with people on the call in a shared virtual […]

This weekend, cinephiles across the world will march to their local theaters to feast their eyes on Christopher Nolan's new adaptation of The Odyssey. It's on track to rake in anywhere between $80-$100 million in just a few days. People are clearly excited to see how Nolan uses cutting-edge filmmaking tech to make the Homeric […]

Neko Health has developed proprietary body-scanning technology, which it couples with bloodwork, to assess a person's health.

OpenAI, which is in the middle of a legal battle with Apple over hardware trade theft allegations, just released a light-up keyboard designed to be paired with its agentic coding app.

Submission for DEV's Summer Bug Smash — Smash Stories track. There was a file in my repo called run_benchmark_1_22.py . Not 1 to 24, which is what the harness was written to do. Not 1 to 26, which is how many Mersenne exponents the agents know. Twenty-two. A chart in the README — a2a_latency_times_1_22.png — agreed. At some point, past-me had decided the benchmark ends at 22, committed the evidence, and moved on. This summer, hunting for a Bug Smash target, I finally asked: why 22? The setup a2a-benchmark compares A2A agent performance across four languages. Python and Go sit behind Gemini tool-calling (ADK); Node and Rust are bare HTTP handlers. Each computes Mersenne primes with Lucas–Lehmer; a harness sweeps N from 1 to 24 and draws two charts. I ran the full sweep. At N=24, the Python column printed N/A . Every other language returned data. There it was — not a decision, a crash , worked around by shortening the run until it stopped hurting. The 4,300-digit wall The Python agent's response at N=24 wasn't even subtle about it: "Exceeds the limit (4300 digits) for integer string conversion; use sys.set_int_max_str_digits() to increase the limit" CPython 3.11 added a default cap on int→str conversion — 4,300 digits — as a denial-of-service mitigation. My agent stringified every prime it found. The 24th Mersenne prime, 2^19937−1, has 6,002 digits . Here's the part that made me laugh out loud: the stringified list was never returned . The tool reports only its elapsed time. The line that had silently amputated my benchmark at N=23 was decorative. The fix was git rm energy: delete the str() , keep the raw int. Go had the identical dead weight ( val.String() ) inside its timed region — it just happened not to crash. One deleted expression, and a column of data that had never existed came into being: N=24, Python, 2,425.9 ms. It gets worse before it gets better With the agents finally running, I kept pulling the thread. The harness parsed Python's elapsed time out of th

You have a project with 20 dependencies. Half of them are outdated. Running ncu -u or pip install --upgrade upgrades all of them at once — and when something breaks, you have no idea which package caused it. So you don't upgrade. The deps rot. Security patches pile up. loopgrade fixes this. It upgrades one dependency at a time, runs your test suite after each upgrade, commits if it passes, reverts if it fails, and moves on to the next one. When it's done, it gives you a full report of what was upgraded, what failed, and why. GitHub: https://github.com/Sagar-S-R/loopgrade PyPI: https://pypi.org/project/loopgrade/ Open for Contributors #Python #LangGraph #opensource

I gave one model the same 44-node architecture twice. The first time I asked for raw SVG — place every box, route every edge, hand me the coordinates. The second time I asked it to describe the same system as typed JSON and let a layout engine draw it. Same model, same session, same brief. The only thing I changed was the output boundary. The boxes are fine in both. I want to be upfront about that, because the usual version of this pitch is out of date. Models place labeled boxes well now. If you ask a current frontier model for a six-box flowchart as SVG you get a clean six-box flowchart, and if that's what you need, go do that — it's the right tool and I'm not going to pretend otherwise. What broke was the edges. With no routing algorithm the model just drew long diagonals straight through unrelated boxes. Not a few — everywhere the graph got dense. And when I changed one node, the entire hand-placed coordinate layout had to be regenerated, and came back different. That second part is the one that actually annoyed me. It's not a rendering bug you can squint past. The picture is a dead artifact: you can't diff it, you can't edit one box, you can't get the same one twice. Every change is a full regeneration and a fresh roll of the dice. This isn't a "wait for a better model" problem Here's the part I'd push back on if someone else wrote it, so let me make the case. Routing a connector around obstacles across a nested graph is global constraint optimization. It's the specific thing layout engines like ELK exist to solve. A model emitting SVG has to commit to an x/y for every point, in order, with no way to backtrack once it sees the whole picture — it's predicting the next token, not solving a layout. So a better model gives you nicer boxes, not untangled edges . The failure is structural, and I'd expect it to reproduce across models past a couple dozen nodes. If you don't buy that, the honest move is to test it: throw a 40-node architecture at whatever model you tru

pinto lets teams manage Scrum backlogs as plain text. Every backlog change can be inspected with git diff , carried across a branch, merged, and reviewed in a pull request—just like source code. The backlog follows the same lifecycle as the product it describes. pinto is local-first: no server, account, or hosted database is required. It works with AI agents, but never depends on them. And unlike a generic CLI to-do list, its core model is Scrum: Product Backlog Items (PBIs), Sprints, Kanban, and the metrics teams use to inspect and adapt. Why pinto exists Jira, Asana, and Notion are capable products. But as more features accumulate, the process can start serving the tool instead of the other way around: Creating one ticket means navigating required fields and workflow settings. Running a Sprint begins with configuring permissions, automations, or dashboards. Even a small status change feels expensive when the tool is slow to open. Scrum is meant to be a lightweight framework for rapid inspection and adaptation . When maintaining the board becomes work in its own right, the tooling has lost sight of that goal. The name pinto comes from the Japanese word ピント “focus.” That is also the design intent: keep the team's attention on the Product Backlog, the Sprint, and the work flowing across the board. pinto deliberately stays small. It starts quickly, keeps dependencies and vocabulary limited, stores durable data as readable files, and excludes project-suite features such as Gantt charts and CRM. Your backlog belongs in the repository Running pinto init creates a .pinto/ directory beside your code. Each PBI is a Markdown file with TOML frontmatter. Here is a real item from pinto's own backlog: +++ id = "P-2" title = "Investigate and fix Windows CI while retaining Windows support" status = "done" rank = "j" labels = ["ci", "windows"] created = "2026-07-15T06:24:58.731847Z" updated = "2026-07-15T09:34:50.653786Z" +++ # Summary Investigate, verify, and fix the current Windo

OpenAI's collaboration with keyboard maker Work Louder is available to order today.

Babies are tremendous learning machines, and key advances for AI may soon be found in the architecture of their little brains.

The stock has steadily fallen from the euphoric post-IPO high, showing that markets may be sobering up to the promises CEO Elon Musk made before and after SpaceX went public.

The stock has steadily fallen from the euphoric post-IPO high, showing that markets may be sobering up to the promises CEO Elon Musk made before and after SpaceX went public.

Inkling, a 975-billion-parameter open source model, was trained to understand video and audio. It could help Thinking Machines establish itself among competitors like Anthropic and OpenAI.