Manji

@OlayaCFK @twentworth12 @SoniaBegh @incident_io Multiple regions. Dozens of countries. Millions of users. "Rest of World" lol😂

@twentworth12 @SoniaBegh @incident_io Does it seem complicated to modify a drop-down menu @twentworth12 ?

Seriously ??? @incident_io

@Kaoutarboutayeb @SoniaBegh Exactly. We should define baselines ""before"" the pilot: what’s ""normal"" for incident categories, handoffs, and resolution paths. Then evaluate AI suggestions for accuracy + usefulness, not vibes. Also: keep humans in the loop early.

@SoniaBegh Client, side reality : buyers don’t want “AI.” They want fewer escalations, clearer ownership, faster resolution. Procurement will ask : where does data go, who can access it, what’s the rollback plan? If we can answer that simply, adoption is smoother.

Worth a read if you’re trying to turn "AI in IT" into actual operations. The promise isn’t magic, it's better signals + better workflows. The risk is adding "another" layer of dashboards. What’s the fastest path to time-to-value without blowing up governance?

@SoniaBegh Most of this has been around for a while, but what’s changing is packaging + operational UX. Still, AI is only as good as telemetry quality. It breaks when logs are inconsistent, labels are missing, or integrations are brittle. Guardrails + eval matter more than demos.

@HElkouzati @MhimNajwa94230 @Kaoutarboutayeb @SoniaBegh Nothing beats the moment systems run smoothly and nobody calls engineering in panic.

@MhimNajwa94230 @Kaoutarboutayeb @SoniaBegh Yes. Calm users usually mean something solid is happening under the hood.

The real moat is operational: consistent infra, controlled data paths, and audit trails teams can actually live with. This is that direction

@SoniaBegh Agentic AI feels real only once observability and guardrails are baked in. Otherwise it’s just confident chaos.

This feels like OpenAI playing long-term chess. More suppliers, more options, less fragility in the stack.

@SoniaBegh Diversifying hardware feels overdue. Betting everything on one supplier turns scaling into a stressful sport.

@SoniaBegh Keeping Slurm open is a relief. Cluster scheduling drama is not something anyone wants locked behind a vendor wall.

Transparency scales trust. Infrastructure scales adoption. Nvidia is tying both together instead of treating them as separate problems.

@DieselNgoma @SoniaBegh Yes, finally a model that doesn’t require turning the office into a mini data center just to add footsteps.

@SoniaBegh The lightweight angle is underrated. Good audio sync without massive compute changes the economics for a lot of teams.

Creators don’t need louder tools, they need tighter timing. Lightweight + well-synced sound is a serious unlock for short-form and product content.

@SoniaBegh This makes AI video feel closer to production-ready and further away from demo-only territory.

Competition is clearly shaping priorities. Tools mature faster once they’re forced to survive real usage, real stakes, real comparaison.

@SoniaBegh This release pace feels intense. Engineers barely finish testing before the next version shows up.

@Kaoutarboutayeb @SoniaBegh Felt, yes… though creators now get so many controls I’m half-expecting someone to request ‘make it sound slightly sarcastic but also hopeful’.

@SoniaBegh The difference between flat audio and expressive audio is basically the difference between "heard" and "felt"

For support, learning and product content, flat narration quietly kills engagement. Giving teams levers for tone and pacing finally brings audio to life.

@hurelbiteny @SoniaBegh Matching maps sounds great, until someone realizes half of us drew ours at 3AM with bad coffee.

@SoniaBegh Standardizing agents feels like cleaning up the wild west of AI. Everyone finally agrees the maps need to match.

Once this many heavyweights agree to standardize instead of just compete, you know something structural is happening.

@SoniaBegh A unified agent ecosystem feels like switching from messy wiring to a clean circuit. Everything just connects.