Whitesmith (@whitesmithco) header
Whitesmith (@whitesmithco) profile picture

Turning AI experimentation into measurable impact

London, UK
Joined May 2013
Photo by Whitesmith (@whitesmithco)Photo by Whitesmith (@whitesmithco)Photo by Whitesmith (@whitesmithco)Photo by Whitesmith (@whitesmithco)Photo by Whitesmith (@whitesmithco)Photo by Whitesmith (@whitesmithco)Photo by Whitesmith (@whitesmithco)Photo by Whitesmith (@whitesmithco)Photo by Whitesmith (@whitesmithco)Photo by Whitesmith (@whitesmithco)
We are hiring! Looking for talented and curious people in Portugal! whitesmith.co/careers/
73
A development team at one of our current clients started sceptical about agents and is now running several cards at once, because an agent takes each specified card from backlog to done, and a developer reviews the screenshots and merge requests. For each card, the agent implements the change, runs the app and the user-flow tests locally and ships to a test environment. Developers say they have less to keep in their heads. When an agent misses something, they report it and update the process themselves. This is one example of the agentic SDLC we build with client teams, where agents take on the steps from spec to test environment, and engineers check their work at each stage.
2
86
Anthropic's Claude Meetings (beta) joins Google Meet as a participant and drafts plans live. OpenAI's dots and xAI's Grok Bot work in your apps from their own cloud computers. On Wednesday 25 November at 4pm Lisbon, in our next webinar, we look at how teams work with AI employees like these and what changes for managers, roles and hiring. whitesmith.co/webinar-leader…
32
AI labs now sell agents as teammates. Before one starts, someone has to decide what it may do alone, what data it can reach and who checks its work. On 25 Nov, 4pm Lisbon, we show how we settle that for our own agents, live. Free. whitesmith.co/webinars/
25
When the developer’s agent writes tests from the spec’s test plan, they tend to restate the acceptance criteria and miss the edge cases a tester would find. We walk through it in “QA in an agentic SDLC”, tomorrow 4pm Lisbon. Free, link in reply.
1
49
This year's last webinars are here! - 11 Nov, for engineering: agents take a user story to an executed test, and a QA engineer reviews every result. 25 Nov, for leaders: our AI employees, each with a role, a manager and limits
1
25
Hiring 2 Forward Deployed Engineers (AI)! Not demo, support or ticket work: you sit with users, build the smallest version that runs the whole workflow, then harden it with evals, tests and human controls. Remote-first, Portugal. To apply: 5-min demo, no cover letter
2
1
1
141
We are hiring talented people who use AI on real work and know where it breaks. Two roles building AI inside real client systems. - Forward Deployed Engineer (AI) - AI Engagement Lead Remote-first in Portugal.
3
1
163
A project manager’s week is full of meetings, status updates, and kicking off new work. Our next webinar will show how AI takes time off each. First, an assistant you check and approve; then an agent that acts on its own, and what it takes to get there. Join us!
27
On a payments feature we shipped, three agents had one job: prove every security finding wrong. Only what survived went to the tracker that gates the deploy. We show the full setup and where it still misfires on 14 Oct. Join us! whitesmith.co/webinar-tech/
1
74
Whitesmith retweeted
AI-pilled in Portugal? I'm looking to hire Engagement Leads and FDEs to work with me. Apply here: whitesmith.co/careers/ Looking for talent, momentum and results over 20 years of experience with Fable 5.5
4
2
1
13
52,785
We moved security review to every commit. It runs at two speeds: deep checks at design time with a named sign-off, and light checks on every pull request. Our colleague Pedro will write about this process and will explain it further in our next webinar. whitesmith.co/blog/From-peri…
60
On a payments feature we shipped, three agents had one job: prove every security finding wrong. Only what survived went to the tracker that gates the deploy. We show the full setup and where it still misfires on 14 Oct.
37
Claude is cool again but codex is still our choice for tool use!
40
We moved security review from a few times a year to every commit. It runs at two speeds: deep checks at design time with a named sign-off, light checks on every pull request. Join us for our next webinar, where we will show our full step-up!
23
One of our clients is spending half a million euros on AI this year, not the way most companies would. It comes down to two things: making people capable and redesigning how the work gets done. They started with a measured pilot, then scaled. Training alone doesn't move the numbers.
43
Save your spot in our next webinar focused on AI!
51