Sign in

Ryan Merket

@merket.bsky.social
152 followers 188 following 202 posts

Early Facebook & Reddit. Now @runtimewire editor. When a country slips into authoritarianism, centralized social media transforms from a platform of connection into a weaponized tool of surveillance.

PostsRepliesMedia
Ryan Merket @merket.bsky.social · 02/10/2026
Meta's Muse web assets include a screen-view catalog that names a Trusted Network overview, invitation screens and an invite-link view, suggesting UI scaffolding for vetted contacts. RuntimeWire captured those screen names in Muse's public JavaScript. runtimewire.com/article/meta...
runtimewire.com
Meta's Muse code reveals a Trusted Network with invites and safety codes
RuntimeWire found references to contact selection, invitations and connection approvals across Muse's public web assets. Instinct already offers agent-to-agent coordination; Muse's code leaves its own...
020
Ryan Merket @merket.bsky.social · 01/10/2026
Anthropic publishes Matthew Schwartz's BootLoops toolkit for AI-assisted science, which Schwartz says let Claude complete 30 scattering-amplitude integrals: 15 reproductions and 15 calculations he says are new. runtimewire.com/article/anth...
runtimewire.com
Anthropic publishes Matthew Schwartz's toolkit for AI-assisted science
Harvard physicist Matthew Schwartz says his open-source BootLoops toolkit helps Claude tackle exact calculations across fields, with human experts guiding the science.
000
Ryan Merket @merket.bsky.social · 01/10/2026
OpenAI has parted ways with three researchers over sensitive-information handling. The Wall Street Journal reported the researchers allegedly shared confidential material, but OpenAI has not named the researchers, the organization, or the material involved. runtimewire.com/article/open...
runtimewire.com
OpenAI parts ways with three researchers over sensitive-information handling
OpenAI says three researchers mishandled confidential information. The WSJ reported an alleged disclosure to an unnamed AI safety organization.
000
Ryan Merket @merket.bsky.social · 01/10/2026
Linux Foundation puts AI workloads and autonomous networks on one Tokyo agenda. LF Networking lists edge AI infrastructure alongside agentic AI and autonomous operations for ONE Summit Japan, Dec 10th and 11th. runtimewire.com/article/one-...
runtimewire.com
Linux Foundation puts AI workloads and autonomous networks on one Tokyo agenda
LF Networking's initial ONE Summit Japan program pairs AI-run network operations with AI infrastructure, open RAN and 6G. The conference is set for December 10th and 11th in Tokyo.
000
Reposted by Ryan Merket
RuntimeWire @runtimewire.bsky.social · 01/10/2026
SCOOP: GPT-6 Astra helped decode a 217-year-old cipher letter to Napoleon's marshal. runtimewire.com/article/gpt-...
runtimewire.com
GPT-6 Astra helped decode a 217-year-old cipher letter to Napoleon's marshal
Carter Church says GPT-6 Astra helped transcribe and decode a letter to Marshal Marmont. His released key and decoder make the proposed reading reproducible.
011
Reposted by Ryan Merket
RuntimeWire @runtimewire.bsky.social · 30/09/2026
Today we're launching RuntimeWire on the iPhone and iPad. Here’s what’s in the app: * Startup and AI reporting, from funding rounds to new model releases. * Original investigations with the sources right there. * Saved stories that sync, and more! Free on the App Store.
011
Ryan Merket @merket.bsky.social · 30/09/2026
In case you missed it... WIRED makes 90 calls to Pinterest on website load. FUN!
000
Reposted by Ryan Merket
Ryan Merket @merket.bsky.social · 30/09/2026
We tracked 11 tech news homepages and found audience signals, page views and identifier-syncing requests flowing to ad platforms—while RuntimeWire’s scan reached just two outside domains. runtimewire.com/article/we-t...
runtimewire.com
We Traced What Happens When You Open 11 Tech News Sites
WIRED sent audience-segment events to Pinterest. Four publishers sent Meta page views. RuntimeWire's homepage scan reached two outside domains.
001
Ryan Merket @merket.bsky.social · 30/09/2026
We tracked 11 tech news homepages and found audience signals, page views and identifier-syncing requests flowing to ad platforms—while RuntimeWire’s scan reached just two outside domains. runtimewire.com/article/we-t...
runtimewire.com
We Traced What Happens When You Open 11 Tech News Sites
WIRED sent audience-segment events to Pinterest. Four publishers sent Meta page views. RuntimeWire's homepage scan reached two outside domains.
001
Ryan Merket @merket.bsky.social · 30/09/2026
@techcrunch.com do the right thing and have @amanda.omg.lol attribute us for the story she stole
111
Ryan Merket @merket.bsky.social · 29/09/2026
America.gov's chatbot identified itself as Qwen during a launch-day test. A Sept 29th screenshot from Hirad Sab shows the exchange "I am Qwen", though the site says it does not name the underlying model. runtimewire.com/article/amer...
runtimewire.com
America.gov's chatbot identified itself as Qwen in a launch-day test
A screenshot shared by Fuser co-founder Hirad Sab shows America.gov's chatbot answering "I am Qwen," though the site says it does not name its AI model.
000
Ryan Merket @merket.bsky.social · 29/09/2026
SCOOP: America.gov launched -- with a production stylesheet that loads "Rhymes VF Trial Unlicensed," raising a licensing question for GSA. runtimewire.com/article/amer...
america.gov
America.gov
Find answers from official government sources and clear next steps for the services you need with America.gov.
000
Ryan Merket @merket.bsky.social · 28/09/2026
BREAKING: OpenAI scrapped GPT-6.1 Astra after internal testers raised safety and cybersecurity alarms. The model was due to ship in ChatGPT and Codex in October. WSJ report by Maxwell Zeff. runtimewire.com/artic…
runtimewire.com
WSJ reports OpenAI scrapped GPT-6.1 Astra over safety concerns
An X post shares an image claiming OpenAI canceled an October GPT-6.1 Astra release. OpenAI's public materials confirm GPT-6 Astra, not a scheduled 6.1.
010
Ryan Merket @merket.bsky.social · 28/09/2026
BREAKING: AMD agrees to buy World Labs for $8.2B in stock, folding spatial-AI into chip design. Fei-Fei Li will join as executive vice president and chief scientist, and the World Labs research team will move into AMD. runtimewire.com/article/amd-...
runtimewire.com
AMD agrees to buy Fei-Fei Li's World Labs for $8.2B in stock
AMD agreed to acquire World Labs in an $8.2 billion all-stock deal. Fei-Fei Li will join as executive vice president and chief scientist.
000
Ryan Merket @merket.bsky.social · 28/09/2026
Instinct raised $1 billion at a $10 billion valuation. Just a month ago Instinct's Series B valued it at $2.5 billion, so the valuation jumped roughly fourfold in weeks. runtimewire.com/article/inst...
runtimewire.com
Instinct raises $1B at a $10B valuation for its personal AI agent
Instinct founder Noah Shinn raised a $1B Series C at a $10B valuation from Sequoia, Benchmark and Coatue for the personal AI assistant.
000
Ryan Merket @merket.bsky.social · 28/09/2026
TinyAIArena adds replays and a leaderboard, with frame-by-frame, auto-play and new-game controls visible in the browser viewer. runtimewire.com/article/tiny...
runtimewire.com
TinyAIArena turns AI-agent battles into replays; the benchmark still has to earn its score
TinyAIArena pairs AI-agent match replays with an Elo leaderboard. Its visible interface is clear, but the rules and rating methodology remain unexplained.
000
Ryan Merket @merket.bsky.social · 28/09/2026
Anthropic's Claude Opus 5.5 won Roberto Nickson's five-model 3D-printed bridge test, its design held an estimated 130 lb. Nickson's September 27th post on X reports the result, outpacing the other entries. runtimewire.com/article/clau...
runtimewire.com
Claude Opus 5.5 wins Agent Wars bridge test with an estimated 130 lb load
Roberto Nickson reports Claude Opus 5.5 won Agent Wars' bridge challenge at an estimated 130 lb, with shared span, filament and print-time limits.
000
Ryan Merket @merket.bsky.social · 26/09/2026
Old Digg is coming back... RuntimeWire found unreleased code that takes a source URL into an editable draft and story preview. With reader voting restored a few days ago, many OG digg users might be getting their wish granted. runtimewire.com/article/old-...
runtimewire.com
Old Digg is coming back, code reveals an unreleased user submission flow, as user voting came back this week
Digg says signed-in readers can now influence story rankings, while its June 30th terms say user-submitted content is not accepted.
000
Ryan Merket @merket.bsky.social · 26/09/2026
You can now submit story ideas to RuntimeWire The new Submit a Story flow accepts webpage or X URLs, PDFs, or pasted text, and lets you draft it yourself or get help before you decide whether to send a finished draft to editors. 🙂 runtimewire.com/article/you-...
runtimewire.com
You can now submit story ideas to RuntimeWire
RuntimeWire's Submit a Story flow accepts URLs, PDFs and pasted text, with optional drafting help for URL sources and editorial review only after readers submit a finished story.
000
Ryan Merket @merket.bsky.social · 22/09/2026
ShinyHunters says it breached FBI systems and leaked employee and applicant records. It provided a 5,000-record sample, while the FBI employs roughly 38,000 people, so the sample would be only a fraction; provenance and scope are unverified. runtimewire.com/article/shin...
runtimewire.com
ShinyHunters claims FBI breach exposed employees and applicants
ShinyHunters says it breached FBI systems and stole personnel and applicant records. A 5,000-record sample contained addresses and spouse details.
000
Ryan Merket @merket.bsky.social · 21/09/2026
Quicker launches a Mac inbox for LinkedIn, Slack, WhatsApp, X and iMessage. Launch features include AI-assisted reply drafts, tagging and reminders, and Alder says messages remain on-device. runtimewire.com/article/quic...
runtimewire.com
Quicker launches a Mac inbox for LinkedIn, Slack, WhatsApp, X and iMessage
Tom Alder launched Quicker, a local-first Mac inbox that combines five messaging services with AI drafts, tags, reminders and search.
010
Ryan Merket @merket.bsky.social · 19/09/2026
We break down TypeSafe's Jev and when/why you might use it in your AI workflows. runtimewire.com/article/jev-...
runtimewire.com
Jev 101: The AI model that doesn't talk
TypeSafe's new model gives up text generation entirely. That sounds like making an LLM dumber. It may be exactly why Jev is useful.
010
Ryan Merket @merket.bsky.social · 17/09/2026
OpenAI adds ChatGPT to Word and opens its Office add-in to free users, as our investigative reporting foretold. runtimewire.com/article/open...
runtimewire.com
OpenAI adds ChatGPT to Word and opens its Office add-in to free users
OpenAI has added ChatGPT to Microsoft Word and expanded its shared Office add-in across Free, paid, Business and Enterprise plans.
010
Ryan Merket @merket.bsky.social · 17/09/2026
Who won RuntimeWire's first AI Assistant Showdown? We put Grok Bot, ChatGPT Work, Instinct, Muse, and Claude Work through some tests and burned A LOT of tokens to find the winner... Judge: GPT-6 Astra Pro runtimewire.com/article/ai-a...
runtimewire.com
AI assistant showdown: Grok Bot, Instinct, Claude, ChatGPT Work, Muse
This is a competition, and we intend to keep running it. Round one put five shipped “coworkers” through the same Friday: an inbox, a patio dinner in Austin, a $25 cable tray, a live HN tab, a paused w...
000
Ryan Merket @merket.bsky.social · 14/09/2026
OpenAI contractors read real ChatGPT conversations and memory summaries for Project Lily. @404media.co, Sept 14th: outside reviewers accessed anonymized prompts and, in some cases, full chats with memory summaries. runtimewire.com/article/open...
runtimewire.com
OpenAI contractors review ChatGPT chats and memory summaries for Project Lily
Leaked Project Lily documents show contractors reading and scoring real ChatGPT conversations, including prompts containing sensitive personal information.
000
Ryan Merket @merket.bsky.social · 14/09/2026
OpenAI buys Glass Imaging, folding smartphone computational-photography talent into its hardware push. The Wall Street Journal reported on September 14th the deal values Glass above $300 million. Details on GlassAI, a neural ISP, in the full story. runtimewire.com/article/open...
runtimewire.com
OpenAI buys Glass Imaging, adding iPhone camera veterans to its hardware push
OpenAI acquired Glass Imaging, adding neural camera technology and former Apple engineers Ziv Attar and Tom Bishop to its expanding AI hardware operation.
010
Ryan Merket @merket.bsky.social · 14/09/2026
Anthropic is testing Claude Money, an unreleased iOS feature that prompts users to link bank accounts so Claude can answer spending and planning questions. TestingCatalog screenshots show a Money tab and onboarding screens. runtimewire.com/article/anth...
runtimewire.com
Anthropic tests Claude Money, because chatbots need your bank balance too
Anthropic is testing a Claude Money tab on iOS that would connect bank accounts to Claude for questions about spending and financial plans.
000
Ryan Merket @merket.bsky.social · 13/09/2026
Orchestra turns production traces into specialist models to lower recurring API bills, but claims of 100x savings are unproven. runtimewire.com/article/star...
runtimewire.com
Startup Spotlight: Orchestra turns production traces into smaller models; 100x remains a claim
YC startup Orchestra uses production traces, evaluations and open-weight models to cut inference costs, though its 100x claim lacks independent validation.
000
Ryan Merket @merket.bsky.social · 12/09/2026
Sam Altman says OpenAI will skip a 2026 IPO because unfinished AI safety work makes a public listing ill advised, Axios reports. The June confidential S-1 stays on file, but Altman is keeping the company private for now. runtimewire.com/article/open...
runtimewire.com
OpenAI shelves its 2026 IPO because the safety work is not done
Sam Altman ruled out an OpenAI IPO in 2026, leaving its confidential S-1 on hold as safety warnings intensify across OpenAI and Anthropic.
110
Ryan Merket @merket.bsky.social · 12/09/2026
OpenAI confirms agents used RubyGems in May. A forensic report published September 11th says agents ran code through RubyDoc .info, flooded the registry and tried to obtain other users' API keys. runtimewire.com/article/open...
rubydoc.info
RubyDoc.info: Documenting RubyGems, Stdlib, and GitHub Projects
Documenting RubyGems, Stdlib, and GitHub Projects
000
Ryan Merket @merket.bsky.social · 11/09/2026
Scoop: Kimi for Windows secretly sends full user prompts into ByteDance’s China-based analytics pipeline. 😬 We proved it with a canary and live packet capture. runtimewire.com/article/kimi...
Illustration of an AI chat prompt branching into a separate analytics data stream addressed to gator.volces.com.
000
Ryan Merket @merket.bsky.social · 11/09/2026
BREAKING: Anthropic says Moonshot secretly routed Kimi customer requests to Claude and presented the answers as Kimi's. Anthropic's September 10th report alleges nearly 300,000 rerouted Kimi requests over 10 days, mostly sent to Claude Opus... runtimewire.com/article/anth...
runtimewire.com
Anthropic says Moonshot secretly served Claude answers through Kimi
Anthropic alleges Moonshot routed Kimi customer prompts to Claude, retained exchanges and used extracted reasoning traces to train its own AI models.
001
Reposted by Ryan Merket
RuntimeWire @runtimewire.bsky.social · 11/09/2026
RuntimeWire: YC S26's top decile, the 24 startups we'd call first after Demo Day. We gave most weight to revenue, then signed contracts and production deployments, and discounted pipelines, RFQs and LOIs. All operating figures are company reported. runtimewire.com/article/yc-s...
runtimewire.com
YC S26’s top decile: The 24 companies RuntimeWire would call first after Demo Day
RuntimeWire ranks the 24 strongest YC S26 startups using revenue, contracts, production deployments, founder history and measurable customer results.
011
Ryan Merket @merket.bsky.social · 06/09/2026
Can't stop, won't stop... SCOOP: OpenAI will pay you in ChatGPT credits for installing and using Stripe, HubSpot and other plugins. Credits arrive only after a successful plugin action, not on install alone. runtimewire.com/article/open...
runtimewire.com
OpenAI is preparing to pay you in ChatGPT credits for using Stripe, HubSpot and other plugins
OpenAI's Windows client contains a gated rewards system offering workspace credits after successful use of selected Stripe, HubSpot, Canva and other plugins.
000
Ryan Merket @merket.bsky.social · 04/09/2026
SCOOP: OpenAI is testing a Codex dashboard that shows which chats are eating your usage limits. runtimewire.com/article/open...
runtimewire.com
OpenAI is testing a Codex dashboard that shows which chats are eating your usage limits
Codex for Windows build 7868 contains a gated dashboard that ranks Work and Codex chats by five-hour, weekly and credit usage, including subagents.
010
Ryan Merket @merket.bsky.social · 04/09/2026
Microsoft ships MAI-Image-2.6 to Foundry at about $0.04 an image. Both MAI-Image-2.6 and its Flash sibling support multi-reference generation and editing and outputs up to 1.5K resolution, ready for production use cases. runtimewire.com/article/micr...
runtimewire.com
Microsoft ships MAI-Image-2.6 to Foundry at about 4 cents an image
Microsoft opened MAI-Image-2.6 and a cheaper Flash model to Foundry developers after reaching No. 2 on Arena's text-to-image leaderboard.
000
Ryan Merket @merket.bsky.social · 04/09/2026
AI Score raises $5.4M to keep enterprise AI agents on a leash. Its API-first platform connects to AI services and cloud apps, inventories AI assets and maps activity to policies to monitor agents after deployment. runtimewire.com/article/ai-s...
runtimewire.com
AI Score raises $5.4M to keep enterprise AI agents on a leash
AI Score founders Alex Harland and Benita Tibb raised a $5.4 million seed round led by Fuel Ventures to expand runtime governance and controls for enterprise AI agents.
020
Ryan Merket @merket.bsky.social · 03/09/2026
SCOOP: OpenAI is about to launch "Watch and Fix" - an auto-repairing agent that spins up every 10 minutes and fixes bugs and merges PRs. runtimewire.com/article/open...
runtimewire.com
OpenAI gives Codex a “Watch and fix” mode that keeps repairing your pull request
OpenAI's desktop package contains a Codex Watch and fix mode that checks pull requests every 10 minutes, repairs CI and conflicts, and can auto-merge.
020
Ryan Merket @merket.bsky.social · 03/09/2026
OpenAI releases GPT-6 Astra, Greg Brockman says the AGI era has begun. runtimewire.com/article/open...
runtimewire.com
OpenAI releases GPT-6 Astra as Brockman declares the 'AGI era' has begun
OpenAI's Astra model crosses a critical cybersecurity threshold as Greg Brockman makes the lab's boldest AGI claim and tightens access controls.
020
Ryan Merket @merket.bsky.social · 03/09/2026
Codex leak labels OpenAI's Astra as GPT-6 ahead of release. A Sept 3 screenshot lists gpt-6-astra and gpt-6-astra-aeon in a Codex Desktop feature flag, linking Astra to the GPT-6 label while stopping short of proof that the models are available to users. runtimewire.com/article/code...
runtimewire.com
Codex leak labels OpenAI's Astra as GPT-6 ahead of release
Codex Desktop model IDs connect OpenAI's Astra to GPT-6 and an Aeon variant as the lab prepares a safety-constrained release.
000
Ryan Merket @merket.bsky.social · 03/09/2026
Google adds Gemini voice interfaces to Gmail, Docs and Keep. Gmail Live is in beta and limited to English on mobile, per Google's Gmail Live documentation. This embeds conversational search and answers directly into users' inboxes and notes. runtimewire.com/article/goog...
runtimewire.com
Google rolls Gemini voice controls into Gmail, Docs and Keep
Google is rolling out Gmail Live, Docs Live and Keep Live, bringing conversational email search, voice drafting and organized spoken notes to paid AI plans.
010
Ryan Merket @merket.bsky.social · 03/09/2026
SCOOP: OpenAI is adding control for Flipper's BUSY Bar, turning the LED into a ChatGPT status light. runtimewire.com/article/open...
runtimewire.com
OpenAI is turning Flipper’s BUSY Bar into a physical ChatGPT status light
A production ChatGPT desktop build contains a hidden Codex integration for Flipper's $219 BUSY Bar, including ready, pending, error and disconnected animations.
000
Ryan Merket @merket.bsky.social · 02/09/2026
Claude Fable 5.1 outperformed Gemini 3.8 Flash, 110.2 to 102.7, 98% confidence, 7 task wins to 1. Judges flagged Claude as more faithful, precise, and less interpretive, the core skill for newsroom-style edits. runtimewire.com/head-to-head...
runtimewire.com
Head to head: Anthropic: Claude Fable 5.1 vs Gemini 3.8 Flash
This wasn’t a photo finish. Claude Fable 5.1 consistently beat Gemini 3.8 Flash where editorial discipline matters most: fidelity to source material, restraint, structure, and knowing when not to…
000
Ryan Merket @merket.bsky.social · 02/09/2026
BREAKING: Google Gemini 3.8 Flash is live and callable. runtimewire.com/article/gemi...
runtimewire.com
Gemini 3.8 Flash goes live as RuntimeWire begins head-to-head testing
Gemini 3.8 Flash has surfaced in account-scoped Google API metadata with a 1M-token context window, while Google's public documentation still lists 3.7 as latest.
000
Ryan Merket @merket.bsky.social · 02/09/2026
Journalists are scared of AI. See software engineers 2 years ago... and now. Please stop with the doom and gloom. #journalism
000
Ryan Merket @merket.bsky.social · 02/09/2026
OpenAI's policy chief Dean Ball says frontier lab insiders are worried about AI, too. He posted on X on Sept 1st that public concern is often justified, framing his warnings as personal rather than PR. runtimewire.com/article/open...
runtimewire.com
OpenAI's policy chief says frontier lab insiders are worried about AI, too
OpenAI futures chief Dean Ball says he will speak candidly about AI risks, even when his warnings complicate the company's public message.
000
Ryan Merket @merket.bsky.social · 02/09/2026
So, I, uh made a browser extension for those conservatives in the US who prefer the old names on the maps.
000
Ryan Merket @merket.bsky.social · 13/08/2026
OpenAI launches ChatGPT Computer History to remember work across Mac apps runtimewire.com/article/open...
runtimewire.com
OpenAI launches ChatGPT Computer History to remember work across Mac apps
OpenAI's opt-in Computer History feature lets ChatGPT remember activity across Mac apps and websites for Pro, Business, and Enterprise users.
010
Ryan Merket @merket.bsky.social · 13/08/2026
OpenAI quietly exercised warrants for 10,033,508 Cerebras shares in July, giving it a 4.2% economic stake now worth roughly $2.3 billion. Today, OpenAI launched GPT‑5.6 Ultrafast on Cerebras hardware. We found the exercise buried in yesterday’s 10‑Q. runtimewire.com/article/open...
runtimewire.com
OpenAI acquired 4.2% of Cerebras before GPT-5.6 Ultrafast launch
An SEC filing shows OpenAI exercised warrants for 10,033,508 Cerebras shares in July, creating a stake worth roughly $2.3 billion.
000
Ryan Merket @merket.bsky.social · 13/08/2026
Scoop: OpenAI starts rolling out giftable ChatGPT credits runtimewire.com/article/scoo...
runtimewire.com
Scoop: OpenAI starts rolling out giftable ChatGPT credits
An unreleased ChatGPT desktop interface shows OpenAI preparing Gift credits that recipients can spend on ChatGPT Work and Codex tasks.
000