Five stories that mattered this week. Curated for people who don’t have time to miss what matters.
Networkcraft Desk · 6 min read
$11B — ElevenLabs valuation
$500M — Series D raise
$330M+ — ElevenLabs ARR
82 — ransomware incidents
31% — healthcare targeted
AI Models
Claude Opus 4.6 Is the New Benchmark Leader
Anthropic’s Claude Opus 4.6 launched February 5 and immediately took the top slot on Terminal-Bench 2.0, Humanity’s Last Exam, and the GDPval-AA composite ranking — leading GPT-5.2 by 144 Elo points. The headline feature is the 1M token context window (beta), enabling full-codebase audits, entire legal document portfolios, and extended agentic sessions without context loss.
Less-covered but arguably more important: agent teams — native multi-Claude orchestration within a single workflow — and context compaction, which automatically summarises stale context to reduce token waste. There’s also a “Claude in PowerPoint” enterprise preview.
Enterprise clients already live: Deutsche Telekom, Revolut, Meta, Salesforce. Anthropic is at $14B ARR growing 10x. Claude Code alone is at $2.5B+ ARR. The catch: 1M-token contexts are expensive at scale.
AI Infrastructure
ElevenLabs Raises $500M at an $11B Valuation
ElevenLabs closed a Sequoia-led Series D at an $11 billion valuation — 3x its valuation from just 13 months ago. Current ARR is $330M+, and the enterprise client list now includes Deutsche Telekom, Revolut, Meta, and Salesforce. Voice AI has moved from novelty to infrastructure.
The velocity here is the real signal. Growing from a sub-$4B valuation to $11B in 13 months, with real ARR and Fortune 500 contracts, is not hype-curve behaviour. This is a voice API becoming a utility layer.
Cybersecurity
Lotus Blossom and the State of State Hacking
A Chinese-attributed threat group called Lotus Blossom hijacked the Notepad++ update channel — turning a trusted software updater into a malware delivery mechanism. Separately, UNC3886 hit a Singapore telecom and Russia is now using generative AI to accelerate exploitation of fresh CVEs, with CVE-2026-21509 weaponised within days of public disclosure.
82 ransomware incidents in February. 31% in healthcare. The pattern is consistent: state actors pre-position, criminal actors monetise, and both are now operating with AI assistance. These are not separate threat landscapes.
Robotics & Hardware
Robot Home Is Real Now
Four hardware signals from this week: Boston Dynamics Atlas is now operational in a Hyundai manufacturing plant — the first confirmed deployment of a bipedal humanoid robot in active production. Roborock has opened pre-orders for its home autonomy robot. LG CLOiD completed its second public demo, showing improved object manipulation. And Pebble Watch 2 has begun shipping to backers.
These four don’t feel related, but they are: the hardware layer of the AI era is arriving simultaneously across industrial, domestic, wearable, and service robotics. It’s not a trend. It’s convergence.
OpenAI
OpenAI Formally Deprecates GPT-4o
February 13 is the formal deprecation date for GPT-4o. OpenAI’s full lineup is now the GPT-5.x era — GPT-5.2 as the flagship, GPT-5.2-Codex for developer workloads, and the o-series for reasoning-heavy tasks. GPT-4o technically still functions past the deprecation date for existing API customers under sunset policy, but it is no longer the recommended default.
If your production systems are still routing to GPT-4o, now is the time to migrate. The performance gap between GPT-4o and GPT-5.2 is significant — and it’s only going to widen as OpenAI focuses engineering resources entirely on the 5.x family.
ElevenLabs at $11B Is a Tell
Everyone is watching LLM benchmarks. But ElevenLabs going from sub-$4B to $11B in 13 months, with $330M ARR and Deutsche Telekom in the client list, tells you something about where enterprise AI spending is actually landing: voice interfaces and audio generation are becoming infrastructure, not features.
The companies treating voice AI as a vertical product — not a bolt-on — are building something with significant switching costs. Once your IVR, your customer service pipeline, your accessibility stack, your multilingual content production workflow runs on ElevenLabs, you don’t casually swap it out.
$11B for a voice API sounds like a lot. In three years, it may look like the early AWS valuation arguments. The infrastructure layer of the AI era doesn’t have to be a model.
Related Reading
Every Sunday — five stories, one opinion, zero filler.