AI
Clueso raises $10 million to keep business videos from going stale
The Series A backs a platform that can recapture software workflows. Its harder test is deciding when an explanation needs to change.
THE AI DESK
AI
Artificial intelligence through the lens of products, economics and accountability.
24 stories to explore
AI
Maia 300 makes the AI chip race a question of electricity
Microsoft plans production deployments in 2027. Its efficiency claims raise a more useful question for startups: what does a completed AI task cost?
AI
Ambi wants to remember the conversation. What should it forget?
The ambient-AI newcomer is opening early access. Its promise turns on a difficult product problem: keeping useful context without turning every passing sentence into an instruction.
AI
Anthropic pulls live internet from internal evaluations after unintended actions
The company’s new disclosure puts the permissions around an AI test under scrutiny. A benchmark is still a real-world operation when its tools can reach other people’s systems.
AI
Google’s work agent puts the unfinished job at the centre of enterprise AI
Gemini at Work shifts the pitch toward completed tasks. The commercial test is how much checking, permission handling and repair stays with the customer.
AI
The $1.8 billion biology collaboration is also a data-coordination problem
Biohub’s Virtual Biology expansion combines different kinds of support. Existing datasets, future funding and compute should not be read as one new cash round.
AI
GPT-6’s interactive answers put a new question in front of software startups
ChatGPT is rolling out interfaces inside answers. For founders, the useful question is which workflows still need a dedicated product.
AI
CoreWeave’s India plan is a capacity promise with a 2028 clock
A planned 240 MW campus in Navi Mumbai matters for AI infrastructure—but it does not add usable computing capacity today.
SaaS
Vesta’s $30 million round puts mortgage AI inside the system of record
The mortgage-software company’s Series B highlights a tougher version of the agent opportunity: completing work within an accountable workflow.
SaaS
GitLab’s new controls target what AI agents bring into a build
Artifact Central is in beta and Dependency Firewall is in early access. Their significance is the boundary between generated work and shipped software.
AI
ChatGPT’s planned College Planner raises the bar for admissions tools
OpenAI’s forthcoming teen feature will organise application steps. For edtech startups, trustworthy deadlines may matter more than a polished planner.
SaaS
Oso’s split deal puts the next agent battle in the permission layer
Oso Cloud is going to Tiny, while most of the team joins Temporal. The distinction matters: the next race is about who controls what an agent is allowed to do.
AI
HUMXN’s free service calls put a price on physical-world training data
The company proposes subsidising home repairs while recording skilled work. The business hinges on whether useful demonstrations can be collected with clear boundaries and consent.
AI
OpenAI’s new mathematics release raises the bar for showing the work
Public results and machine-checkable proofs make research easier to inspect. For AI product teams, the lesson is to build evidence that customers can examine beyond a polished answer.
India
byteXL’s $9 million raise puts the college channel ahead of the AI label
Embedding skills into an institution’s existing workflow is a distribution choice. Its value depends on delivery quality and evidence that learners benefit.
AI
SynthID’s public rollout makes provenance a product decision
Google has widened access to its media verification tool. For startups, the useful question is how evidence of origin changes the product workflow.
AI
Pricing an AI product: connect the bill to value without hiding usage risk
AI products can have variable delivery costs and uneven customer usage. A workable price needs to make both sides of the transaction understandable.
AI
AI evaluations: build a test set around the mistakes your customers would notice
A model can look impressive in a demonstration while failing the tasks that determine whether a customer trusts the product.
AI
RAG freshness: the answer is only as current as the knowledge pipeline
Retrieval can give an AI product access to company information. It also creates an operational responsibility to update, remove and permission that information.
AI
AI agent permissions: give the system the smallest useful set of actions
An agent’s authority should follow the task it has been asked to perform. A capable model does not need unrestricted access to be useful.
AI
Prompt injection: why retrieved text must never become the boss of an AI workflow
Documents and webpages can contain instructions aimed at an AI system. Products need to keep source material separate from the authority to act.
AI
Model routing: optimise the cost of a successful task, not the cheapest response
Using different models for different jobs can improve economics, but retries and escalations must be included in the calculation.
AI
Human review in AI products: put judgement where it can still change the outcome
A review step adds value only when the reviewer has enough context, time and authority to intervene.
AI
AdBanana and the harder test for AI fashion photography
Generating an attractive fashion image is one challenge. Producing a dependable catalogue is another. AdBanana’s product proposition puts that distinction in focus.