Tools
A coding tool gains an intensive execution mode and dynamic workflows
The features target continuous work rather than question and answer.
Billing split because agents don't consume like conversations
A company created dedicated monthly credits for agent builders, shipped a continuous execution mode and published a high score on an interface operation test. All three point to more autonomous use.
Business
Separate monthly credits for agent builders take effect in June
Agent use spends differently, and charging from one limit penalised the advanced case.
From the studio
More autonomy returns the same question
Three announcements from late May describe the same phase change for anyone working with agents: separate billing credits, a continuous execution mode, and a high score on a graphical interface operation test.
The separate credits are the most revealing market data point. Companies don't split billing for elegance: they split when one group's consumption distorts the general plan. Anyone running agents fires calls in sequence and exhausts the limit doing exactly the work they bought the tool for.
The other two announcements push toward more autonomous use, with the system working until the task is complete and operating applications rather than only writing code.
The counterpart is always the same and has to be decided before switching on automatic mode: what the agent can't do without confirmation, what its reach is inside the system, and what gets logged to allow review afterwards. None of those three decisions depends on a supplier or a budget.
In this issue
AI
Computer-use benchmarks join coding in the scorecard
The 83.4% figure comes from a test requiring operation of a graphical interface, not just writing code.
Business
A $965 billion valuation surpasses the competitor and comes with an IPO filing
The sector's two largest moved toward listing almost simultaneously.
AI
A model reaches 57.9% on a test considered hard, using tools
The detail matters: the number holds with tool access, not in a closed exam.
Work
Continuous agent work requires logging to stay reviewable
What the system does while nobody watches has to be recorded.
Business
The two largest head to market and the sector gains verifiable numbers
Going public turns cost estimates into public information.
Tools
Large cloud platforms prepare their answer in assisted coding
The category is worth more than its own market because it decides where code runs.
AI
Coding numbers rise five points between versions
The gain from 64.3% to 69.2% describes the category's current pace.
Work
Delegation requires defining in advance what the system can't do alone
Continuous execution transfers decisions, and the exception list has to exist.
Business
Agent consumption grew enough to become its own billing line
Creating dedicated credits indicates the usage pattern genuinely changed.
AI
Agents operating computers still fail where the interface changes
A high benchmark in a controlled environment doesn't describe what happens with shifting screens.
Work
A tool that works alone changes what teams need to review
Review designed for hand-written code doesn't describe work done in the background.
Business
Private valuations report market appetite rather than company value
A number negotiated between investor and recipient describes conditions, not fair price.
Tools
The coding tool contest decides where the cloud gets contracted
Choosing the assistant drags execution, storage and billing along.
Work
What the month leaves for anyone putting agents into production
Separate credits, continuous execution and interface benchmarks describe one phase change.
Sections