Reflections on the Beyond the Prompt 2026 conference
Bryntum co-hosted Beyond the Prompt with AG Grid at IET London: Savoy Place on 19 May 2026. Here’s what made a lasting impression.
Beyond the Prompt was honest about AI
The second thing the day got right was the tone; the first was the choice of venue. Most AI discussions are either rose-colored AI hype or “AI will eat your job” doomerism. Beyond the Prompt was a refreshing dose of what’s exciting about AI and where we need to pay close attention.
Speakers across very different companies and roles kept arriving at the same balanced place. AI is here, it’s useful, and to get value out of it, you have to do real engineering work. The talks were also super fun. No one was working on the side, the auditorium remained full for each session, and people enjoyed the live demos, including the TLDraw fairies.
The recognizable AI confidence drop
You often start with high confidence in AI when your agent builds you a slick prototype. And then, as you start actually using the app or running tests, you uncover bugs and strange coding choices. The longer you spend debugging, the lower your confidence in AI drops.
“AI is almost right, but not quite” was a refrain across many talks. That’s why, at its current level of coding ability, it’s best practice to automate what you understand well.
Use AI strategically and with oversight
The talks were different, but the principles overlapped. Bryntum CEO Mats offered the following tips that came up across other talks too.
- Automate what you understand.
- Force evidence. Especially tests.
- If the agent can’t explain it, don’t ship it.
- Keep humans in the loop where risk is high.
- Value over perfection.
Agents are good at executing problems you’ve already solved using approaches and technologies you already understand. If you can’t tell when the output is wrong, or fix the cracks when you find them, you’re left with an unusable end result, plus wasted time and tokens.
So keep humans in the loop. Real users browse the site, a bug surfaces, a GitHub issue gets auto-created, an agent picks it up and opens a PR with a fix and tests, and humans review. That way you benefit from the speed of agents while ensuring code quality.
That bar goes up in critical systems like payments, SDKs, auth, and health, where you have to own every line of code that ships. Agents can help draft and refactor, but human review is non-negotiable for critical-path code.
Forcing evidence showed up in concrete tooling, too. A practical talk on Chrome DevTools MCP demonstrated AI-driven CSS performance debugging, where the agent reads real browser metrics rather than guessing. When the agent can be plausibly wrong, you want the rest of the system feeding it ground truth.
We’re building the tools we need
Engineers are now building the tools they’ve always wanted, like markdown editors, Kanban boards, internal developer utilities, and scripts.
Seeing long-held ideas finally come to life is energizing, and it sparks new ones.
We spoke about the future
On the Bryntum side, our roadmap for the AI era has started. We already have MCP server access to our docs, AI-powered components, and AI chat in our documentation. In the near future, it’ll be even easier and faster for agents to use Bryntum components.
We also used the event to announce that a Bryntum community edition is on the way. More on that in a future post.
A chance to connect in person
Connecting in person was another highlight of the event. We got to have real conversations with clients who’ve been with us for over a decade.
AI means more time in front of the screen, and it does nothing to curtail the Zoom and Teams mode of contact. If anything, it pushes more of our day there.
Dinner after the event gave us a refreshing chance to sit down with the people behind the screens. Catching up across a table is just different. Meeting in person is unparalleled.
Final thoughts
The Beyond the Prompt event cut through the AI noise with grounded talks. To say that AI has changed development is an understatement. But AI has also moved our clients’ needs and expectations. Clients are asking for components that agents can drive, documentation that agents can parse, and faster turnaround on tickets and releases as their own pace picks up.
Bryntum and AG Grid are committed to meeting those needs and exceeding expectations. AI has energized us and affirms more than ever our focus on quality at the center of everything we build.
Massive thanks to AG Grid for hosting alongside us, every speaker for showing up with fun and informative material, Phil Hawksworth for MCing, Sylwia Vargas for organizing the event, and Miguel Salvador for photos of the presentations.