Editorial intro:
The day's top threads share a theme: modern tech creates big, fast signals — revenue claims, calendar events, connected‑car telemetry — but those signals can be unreliable, weaponized, or outright dangerous. Below: quick takes on three stories you should know, then deeper looks at the two that matter most for investors, security teams, and anyone who clicks calendar links.
In Brief
Anthropic are not going to be profitable
Why this matters now: Anthropic’s valuation and run‑rate claims matter for investors and enterprise customers as the company seeks to translate Claude sales into a durable, profitable business.
Reddit users flagged a stark mismatch between headline revenues and underlying losses for AI lab Anthropic. The post cites figures that are now common in coverage: huge top‑line claims paired with massive cash burn.
"Anthropic lost $5.3 billion in 2024 on revenues of $918 million," the Reddit thread notes, and the lab has offered changing run‑rate figures — from $14 billion in early 2026 to more than $30 billion a few months later.
The core takeaway: big revenue numbers don't automatically mean sustainable profit. High inference costs, customer concentration, cloud compute pricing, and political or contractual risks (for example, around defense deals) could quickly erode margin assumptions.
Unsolicited calendar invites are becoming a scammer's favorite way in
Why this matters now: Calendar invites can bypass inbox filters and repeatedly nudge victims, making them a scalable phishing vector for stealing credentials or installing malware.
Security teams are seeing a spike in calendar‑based scams, according to reporting by Business Insider. Attackers use fake meeting requests, malicious .ics attachments, and now AI‑personalization to make invites feel legitimate.
"Calendar-led attacks surged more than tenfold from July," one security analysis cited.
Quick defense: disable auto‑add of calendar events from unknown senders, verify unexpected invites by separate channels, and treat calendar links like any suspicious email.
Chinese cars in Norway stored military exercise footage and access codes
Why this matters now: Connected rental cars captured images and operational details of Norwegian troops, exposing how civilian tech can leak classified or sensitive military data.
Norwegian soldiers found that rental electric vehicles reportedly recorded video and telemetry from a September exercise, including footage of troop movements and, according to reporting, "access codes entered to open military gates and doors" (Militarnyi report). The episode is a reminder that every modern vehicle is effectively a sensor platform; procurement choices now carry data‑security consequences for defense operations.
Deep Dive
Anthropic are not going to be profitable
Why this matters now: Anthropic's revenue and profitability trajectory will influence cloud compute markets, AI pricing, and investor sentiment ahead of any public offering or major enterprise contract rollouts.
The headline numbers are attention‑grabbing: aggressive run‑rate claims and big enterprise deals. But looking closer, the economics of a modern inference‑first AI vendor are tricky. Training a large model is one‑time (but huge); running billions of inference tokens is recurring and scales with customer usage. If Anthropic's contracts are structured with generous discounts, heavy enterprise discounts, or cloud pass‑throughs, the firm’s gross margins can look healthy in presentations yet be fragile in practice.
A few specific pressure points to watch:
- Compute pricing volatility. If Anthropic relies on public cloud instances and market prices rise (or preferred discounts lapse), margins compress fast. Long‑term margin improvement often depends on mix (higher‑margin fine‑tuned or private deployments) and cheaper inference hardware.
- Customer concentration. Enterprise AI deals can be large but brittle. Losing or renegotiating a few major contracts could swing run‑rates dramatically.
- Policy and defense exposure. Ties to defense or national‑security customers bring revenue but also political risk; restrictions or protests could complicate sales and tech partnerships.
Community skepticism on Reddit mirrors these structural concerns: many commenters called the valuation "frothy" and pointed out that optimistic margin slides assume continued scale and favorable cloud economics that aren't guaranteed.
"Big revenue numbers don’t automatically translate into a business that can sustain profits once compute costs, customer concentration and hostile legal or government actions are taken into account," the thread argues.
For investors and enterprise buyers, the prudent move is scenario planning: stress‑test models for higher compute costs, shorter customer lifetimes, and tougher regulatory oversight. For Anthropic, durable profitability will likely come from a mix of product diversification (higher‑margin private deployments and tooling), verticalized SLAs with better pricing power, and — if possible — owning more of their inference stack to reduce third‑party cloud exposure.
Key takeaway: high run‑rates are signals, not guarantees. Anyone valuing Anthropic should model the downside as aggressively as the upside.
Unsolicited calendar invites are becoming a scammer's favorite way in
Why this matters now: Organizations and individuals are starting new quarters and many calendars are full; attackers exploit that normalcy to increase success rates with phishing and credential theft.
This vector works because calendars sit slightly outside the normal inbox workflow. Many clients add events automatically, reminders look like action items, and repeated notifications create a sense of legitimacy. Scammers exploit those affordances with a few tactics that security teams need to block:
- Malicious .ics attachments and links that lead to credential‑harvesting pages or invite recipients to join a scam video call.
- OAuth‑style redirect flows embedded in invite links that request access permissions and can silently steal tokens or grant persistent access.
- AI‑personalized invites that use public bio data or recent email threads to craft believable call titles and agendas.
A short aside on OAuth risks (one specialized but important concept): attackers increasingly try to get victims to click legitimate consent screens that ask for calendar or mail access. When users approve, the attacker may gain ongoing access without ever needing a password. That’s why blocking unknown consent requests and auditing third‑party app permissions is critical.
Operational protections are straightforward and effective:
- Disable auto‑add/new‑event autofill from untrusted senders.
- Train staff to verify unexpected high‑impact invites by a second channel (DM, phone, or internal directory lookup).
- Harden identity controls: conditional access, MFA, and a low‑privilege posture for apps requesting calendar scopes.
On the user side, behavioral changes are low‑cost and high‑impact: treat every calendar invite that asks you to click, upload, or approve as you would a suspicious email. For security teams, logs from calendar APIs are now a must‑have telemetry source; they show who was invited, who accepted, and what links were included — crucial for investigating lateral movement after a successful phish.
"When Pete Thomas spots a meeting on his calendar that he doesn't remember scheduling, the startup executive launches a mini investigation," the Business Insider piece notes — that kind of skepticism should be standard operating procedure.
Closing Thought
Connectedness is a double‑edged sword: it accelerates business models and convenience, but it also amplifies fragility. Whether you're pricing future AI profits, locking down enterprise calendars, or vetting which rental fleet your troops use, treat the flashy signal (run‑rates, calendar reminders, recorded video) as the start of an investigation, not the final answer.