/
Navigation
Chronicles
Browse all articles
Explore
Semantic exploration
Research
Entity momentum
Nexus
Correlations & relationships
Story Arc
Topic evolution
Drift Map
Semantic trajectory animation
Posts
Analysis & commentary
Pulse API
Tech news intelligence API
Browse
Entities
Companies, people, products, technologies
Domains
Browse by publication source
Handles
Browse by social media handle
Detection
Concept Search
Semantic similarity search
High Impact Stories
Top coverage by position
Sentiment Analysis
Positive/negative coverage
Anomaly Detection
Unusual coverage patterns
Analysis
Rivalry Report
Compare two entities head-to-head
Semantic Pivots
Narrative discontinuities
Crisis Response
Event recovery patterns
Connected
Search: /
Command: ⌘K
Embeddings: large
TEXXR

Chronicles

The story behind the story

days · browse · Enter similar · o open

Hark, founded by Figure AI CEO Brett Adcock, previews Handoff, a computer use agent it says outperforms GPT-5.4 and Opus 4.8, and plans for a summer release

TechCrunch Ivan Mehta

Context & Ripple Effects

Hark enters a computer-use-agent race that OpenAI helped define with ChatGPT Agent's ability to control a computer and complete multi-step work. OpenAI has since broadened that work-oriented pitch through ChatGPT Work's access to context across apps and files.

Handoff matters because Hark is publicly framing its forthcoming product against named frontier models, making computer-use performance—not just conversational ability—a focal competitive claim.

First-order effects

  • Hark's performance claim puts GPT-5.4 and Opus 4.8 in an explicit computer-use comparison, while Hark must convert its preview into a summer product release.
  • Prospective users gain another announced option for multi-step computer work, but cannot evaluate Handoff's claimed lead until the product is available.

Second-order effects

  • OpenAI's ChatGPT Work will be compared not only on cross-app context but also on the computer-use performance Hark has made central to its launch positioning.
  • Anthropic and OpenAI face more pressure to substantiate agent capability through task execution as competitors use their models as public benchmarks.

Third-order effects

  • If new entrants can credibly differentiate on computer-use execution, workplace-agent competition will increasingly center on agents as operating layers for tasks rather than standalone chat interfaces.
  • The market's durable dividing line may shift toward verifiable task completion across software environments, with benchmark claims requiring product availability to carry weight.

The trend: AI assistants are evolving into agentic work surfaces whose competitive value is defined by executing multi-step work across existing software.

Discussion

  • @adcock_brett Brett Adcock on x
    Today we're introducing Hark Handoff Handoff has been independently verified as the best internet-use model ever built, outperforming ChatGPT 5.4 & Opus 4.8 While others focus on coding, we focus on everyday life: ordering food, booking flights, shopping, & navigating the web [vi…
  • @hark_labs @hark_labs on x
    Introducing Hark Handoff It is verifiably the best model ever built for using the internet. It can order food, book flights, shop, deeply research, you name it - better and more affordably than frontier models. [video]
  • @chamath Chamath Palihapitiya on x
    Pretty cool.
  • @adcock_brett Brett Adcock on x
    Handoff is a fully capable environment that can handle the open-ended, unpredictable nature of the web to complete end-to-end tasks [video]
  • @chrisgpt Chris on x
    According to Hark's numbers, Handoff costs $0.51 per million input tokens, $9.09 per million output tokens and takes 3.0 seconds of model latency per turn. Compared with GPT-5.5: 10x cheaper input 3.3x cheaper output 2.3x lower model latency per turn I still think [image]
  • @adcock_brett Brett Adcock on x
    Claiming the top spot on the industry's most well-known online browser-use eval, Handoff demonstrates frontier-level performance across three browser-use benchmarks [image]