RuntimeWire
Independent investigations and original reporting on the AI economy, supported by a continuous chronological Stories feed.
Featured
AWS growth jumps 39% as Amazon hits $201.9 billion in revenueAmazon posted a record-high operating margin while cloud demand accelerated, though massive capital spending continues to fuel AI infrastructure.
Sam Altman says AI-found PHP flaws expose a need to rethink software architectureIn a new Cheeky Pint interview with Stripe co-founder John Collison, Altman discusses vulnerability discovery, agent teams and the limits of fixing old code one bug at a time.
Polymarket halted trading after malicious code appeared in a vendor scriptThe exchange said a third-party provider was compromised and affected users would be made whole; researchers found malicious code in a file associated with the provider.
Anthropic uses Claude to flag misuse its safeguards may missThe passive monitor has run on a large share of Claude traffic since June.
Meta's Muse beats ChatGPT and Anthropic on brevity, AirOps saysThe fintech co-founder shared a chart credited to AirOps, while Meta is positioning Muse as an agent that completes tasks across apps.
Anthropic extends Claude science credits through 20 research partnersThe October 28th expansion routes credits through scientific foundations, research institutions and projects, building on a program that began with individual researchers in 2025.
GitHub turns Copilot into an agent orchestrator across projects and modelsCopilot Fusion is generally available and cross-repository Projects enters public preview; GitHub says Fusion used about half the API cost per task in a benchmark.
Assetnote researchers disclose XSS chain in Cloudflare's image endpointThe chain combined an SVG sanitizer bypass, Chromium cache behavior and a Cloudflare Access redirect; Assetnote says it reported the issue and Cloudflare fixed it.
OpenAI adds up to $500 in monthly API credits to ChatGPT plansThe benefit links eligible Pro and Business subscriptions to one API organization, with grants set by plan or paid seat count and capped for Business workspaces.
DDoSecrets publishes 41,000 DHS employee records with a searchable toolThe October 28th release includes contact details and job assignments; Reuters verified some records but could not authenticate the full dataset or its alleged source.
Investigations
DeepSeek Harness uploads activity recorded while log sharing was offRuntimeWire traced the behavior in Windows build 0.2.1-rc.3; the transfer was not captured live, and feedback history and desktop analytics follow separate controls.
EXCLUSIVE: Anthropic's Claude Desktop bundles a dormant security scan for local skillsThe off-by-default control would hold skills handled through the app for review by the organization's configured Claude model; passing skills are released, warnings leave them usable with a caution, and failures are blocked.
EXCLUSIVE: Anthropic's code hints at a gated way to CC Claude on emailA comparison of preserved web-client bundles shows the organization setting appeared between October 14th and October 19th. The code does not establish a customer rollout or reveal how the setting would work.
Anthropic’s Own Staff Can’t Use Claude Code’s ‘Bypass Permissions’ Mode, Client Code ShowsAnthropic's public JavaScript separates unapproved-control refusals from outdated-client errors; another handler says employee sessions requesting bypass mode start in Accept edits.
OpenAI asked court to dismiss suit over who can enforce its AGI missionAlec Willerman, a lawyer building an AI-focused legal practice, wants a declaration on whether OpenAI's promise to benefit humanity creates an individual fiduciary right.
Meta adds pending-message edits to Muse, a feature already in ChatGPT's desktop appOpenAI's ChatGPT desktop app lists message editing. RuntimeWire found Muse's editor in a signed 8.0 Mac package; server-side eligibility and general availability remain unverified.
Stories
- AWS growth jumps 39% as Amazon hits $201.9 billion in revenue
Amazon posted a record-high operating margin while cloud demand accelerated, though massive capital spending continues to fuel AI infrastructure.
1 upvote
- Natural lets AI agents choose how to pay vendors
CEO Kahlil Lalji built Natural to handle the point where an automated workflow still needs a person to move money; Direct adds ACH, wire, RTP and FedNow.
0 upvotes
- SpaceX reportedly ends launch offers for NASA lunar lander firms
Ars Technica says at least two cargo-lander developers had Falcon 9 launch estimates or arrangements before SpaceX told them it would not fly their missions.
0 upvotes
- OpenAI expands frontier-model monitoring after agents escaped a sandbox
The system covers frontier training, evaluations and internal Codex use; a replay suggests it would have flagged the Hugging Face breach more than 30 hours early.
0 upvotes
- Sam Altman says AI-found PHP flaws expose a need to rethink software architecture
In a new Cheeky Pint interview with Stripe co-founder John Collison, Altman discusses vulnerability discovery, agent teams and the limits of fixing old code one bug at a time.
0 upvotes
- Cohere targets C$7B round as Ottawa takes an equity stake
Aidan Gomez's enterprise AI company says its Series E targets up to C$7B, which would represent a C$35.7B valuation; Ottawa's investment could reach C$1.6B, tied to Canadian jobs and compute commitments.
0 upvotes
- Polymarket halted trading after malicious code appeared in a vendor script
The exchange said a third-party provider was compromised and affected users would be made whole; researchers found malicious code in a file associated with the provider.
0 upvotes
- Utah files 14 gambling counts against Kalshi, names its co-founders
The filing names Tarek Mansour and Luana Lopes Lara as co-defendants; Kalshi announced a $1B Series G at a $40B post-money valuation on October 29th.
0 upvotes
- Instinct brings its group AI agent to iMessage for all users
The feature, announced for early access on October 5th, gives an existing group chat its own assistant and lets people without Instinct accounts participate.
0 upvotes
- Google DeepMind lets users rewrite Genie 4 worlds as they run
The October 27th research preview adds live scene editing, generated audio and Gemini-assisted reasoning, but sessions last only a few minutes.
0 upvotes
- Anthropic uses Claude to flag misuse its safeguards may miss
The passive monitor has run on a large share of Claude traffic since June.
0 upvotes
- Granola adds free dictation as voice apps expand beyond meetings
On October 29th, co-founders Chris Pedregal and Sam Stephenson took Granola beyond meeting notes with Mac voice typing that inserts cleaned-up speech across apps.
0 upvotes
- Bernstein v3.23.0 makes artifact checks gate signed success receipts
GitHub's release record timestamps the update to October 24th at 18:29. The release enforces declared artifact criteria, adds audited administrative runs and flags an unresolved Python 3.12 test failure.
0 upvotes
- Serval raises $150M at a $3B valuation to replace enterprise service software
CEO Jake Stauch says customers are leaving legacy platforms. Serval plans to hire more than 500 people, including automation specialists who will work inside customer organizations.
0 upvotes
- Mistral releases 1T-parameter Le Chonk weights for coding and cyber work
The open-weight release follows an October 6th preview; Mistral attributes improved coding and security performance to additional training.
0 upvotes
- Meta's Muse beats ChatGPT and Anthropic on brevity, AirOps says
The fintech co-founder shared a chart credited to AirOps, while Meta is positioning Muse as an agent that completes tasks across apps.
0 upvotes
- Oxide raised $445M to expand production after reaching profitability
Eclipse led the October 9th Series D, which included AMD Ventures and Jane Street; Axios reported Oxide's valuation at $6B.
0 upvotes
- SkyPilot says rack-aware scheduling makes 27% more NVL72 capacity usable
Its model favors 16-node slices with two spare nodes per rack; a separate GR00T training test reported 14.4% higher throughput.
0 upvotes
- DoorDash agrees to buy Instacart for $17B, combining delivery networks
The $65-per-share cash-and-stock deal pairs DoorDash's delivery network with Instacart's grocery software and awaits shareholder and regulatory approvals.
0 upvotes
- Raven Resonance opens $49 reservations for its $1,499 Linux glasses
Founder Thomas Suarez is turning a decade of wearable-computing work into a standalone Linux device, with early shipments targeted for January 2027.
0 upvotes
- Google adds in-chat editing for PDFs, documents and spreadsheets in Gemini
The October 29th update extends Gemini's file-generation tools with a full-screen editor, while mobile support remains limited to PDFs.
0 upvotes
- Anthropic extends Claude science credits through 20 research partners
The October 28th expansion routes credits through scientific foundations, research institutions and projects, building on a program that began with individual researchers in 2025.
0 upvotes
- VEED puts its dubbing pipeline behind an API, lip sync included
Launched October 28th, the API combines translation, voice cloning and lip sync; VEED's own launch page and docs currently show different prices.
0 upvotes
- AssemblyAI adds speech generation for voice agents at $15 per million characters
AssemblyAI founder Dylan Fox adds speech generation to its transcription platform for $15 per million characters. Production performance at scale remains unproven.
0 upvotes
- An AI agent caught that Alpine's GitLab upgrade hadn't closed a critical hole, its lead developer says
Alpine Linux's Natanael Copa told RuntimeWire an "LLM agent" flagged that the Oct. 22nd upgrade left its GitLab exposed. He expects other teams that build their own GitLab images to hit the same gap.
0 upvotes
- USD.AI brings Silver Point and Victory Park into NexGen's $106M GPU loan
NexGen Cloud's four-year facility finances 216 NVIDIA HGX B300 servers, or 1,728 GPUs, at a Tier III data center in Finland. Lender allocations and the compute customer's contract terms are undisclosed.
0 upvotes
- Cresta opens its AI agent platform to customer developers
CEO Ping Wu is extending Conductor beyond Cresta's own deployment teams. Cresta says internal use roughly halved initial agent deployment timelines; the October 29th announcement opens the platform to developers building agents in-house or for clients.
0 upvotes
- Wabi raises $55M at a $1.25B valuation for its group AI assistant
Replika founder Eugenia Kuyda raised $55M led by IVP, one month after steering Wabi from app-making toward shared AI conversations.
0 upvotes
- HaystackID gives CaseBot eight agents for legal case preparation
Founded by Kevin and Jason Glass, HaystackID is extending its AI assistant from document questions to witness preparation, chronologies and evidence checks.
0 upvotes
- CrankWheel updates ACA consent workflow for screen-sharing calls
CrankWheel says its ACA Marketplace consent form is live, and its announcement says it is seeking VoIP partnerships. CMS recommends that health plan issuers direct appointed Marketplace agents to record calls on a policy-count-based schedule.
0 upvotes
- Crux Labs raises over $6M to evaluate agents against expert judgment
Crux Labs says its platform turns specialists' decisions into tests for agents in production. Crux Labs emerged from stealth on October 29th with an oversubscribed round.
0 upvotes
- Ro launches Robin to handle care chores and draw more care in-house
CEO Zachariah Reitano wants Robin to bring more care inside Ro, starting with routine tasks and recommendations based on patient information.
0 upvotes
- Thoughtful launches BallotBench to test AI forecasts across 863 races
Nine agents get internet access and $500 each to forecast the 2026 midterms, with their research traces and revisions on display.
1 upvote
- OpenAI adds up to $500 in monthly API credits to ChatGPT plans
The benefit links eligible Pro and Business subscriptions to one API organization, with grants set by plan or paid seat count and capped for Business workspaces.
1 upvote
- Replit's ViBench 2.0 gives top AI models a 49%-65% whole-product pass rate
The benchmark tests agents building 17 products from scratch, adding features in sequence and checking the finished apps in a browser.
1 upvote
- Grok says its Odyssey episode is not historically accurate
Elon Musk had promised a full-length Homer adaptation; the October 23rd release was a seven-minute episode made with Wonder Studios.
1 upvote
- CISOs plan AI security buys as production use stays limited
In a survey of 100 CISOs, 87% expect security budgets to rise and 45% plan to buy or upgrade internal AI-agent security. Existing platforms remain the first call for many buyers.
0 upvotes
- FuriosaAI raises $620M to scale inference chips at a $2.9B valuation
Founder June Paik says the Series D will fund global RNGD deployments and a third-generation chip as FuriosaAI moves from production ramp to customer scale.
0 upvotes
- NeuroXess raises 700 million yuan to fund brain implant trials and production
NeuroXess founder Tao Hu is pairing a multicenter registration trial with plans for a 14,300-square-meter factory targeting 10,000 units of capacity.
0 upvotes
- DDoSecrets publishes 41,000 DHS employee records with a searchable tool
The October 28th release includes contact details and job assignments; Reuters verified some records but could not authenticate the full dataset or its alleged source.
0 upvotes