OpenAI DevDay 2025 Highlights: All Sessions Now Available

This year’s OpenAI DevDay left me genuinely impressed—yes, I know, not an easy feat in the current landscape of AI conferences. I’m writing this just a few hours after having binge-watched the last available session on OpenAI’s official YouTube playlist. If you’ve missed the live buzz or, like me, enjoy returning to recorded sessions at odd hours with a mug of tea, you’re in luck. Every session is now live to replay, and frankly, there are gems worth a second look.

Let’s slow things down and go through the heart of this year’s event—the breakthroughs, the candid moments, and what all this means for real-world developers, businesses, and AI enthusiasts.

The Next Generation of Tools for AI Creators

I’ve worked with and around machine learning for a good while now, and rarely have I seen so many developer-focused announcements packed into a single conference. OpenAI shed light on its evolving toolkit, making it unequivocally clear—if you’re building on AI, the bar is moving higher, and the tools are getting more accessible.

Here’s a rundown of what made the strongest impression on me this year:

• ChatGPT Apps – A series of updates practically redefining how you and I will use ChatGPT day-to-day. You can now interact with apps inside ChatGPT, and build your own tools with their Apps SDK. Imagine chatting with your calendar, spreadsheets, or even custom workflows, all within a single conversational window.
• AgentKit – This toolkit stands out for anyone who’s ever banged their head against the wall trying to get a “production-grade” AI agent up and running. I found the onboarding shockingly smooth. Gone are the days of debugging endless YAML files; you can now spin up robust assistants with minimal fuss and tune them for real tasks.
• Sora 2 Video Model API – As someone who dabbles with multimedia automation, I was particularly excited by the launch of Sora 2 through API access. The ability to generate and embed video content right from your backend workflow is, well, a big step forward for content creators and app developers alike.
• Codex Updates – OpenAI’s coding model saw nifty upgrades. There’s now tight Slack integration (a godsend for hybrid and remote teams), a refreshed SDK, and improved organisational features. Even my teammates—some who’ve been sceptical—have started poking around with the new workflow features.
• GPT-5 Pro in the API – I got my hands on GPT-5 Pro during the sandbox demo and, honestly, it’s a leap both in speed and comprehension. The improvements felt tangible, especially when handling nuanced, business-critical tasks. If you’ve run into “AI fatigue,” this might be the boost you’ve been waiting for.
• Mini Models: gpt-realtime-mini & gpt-image-1-mini – These lighter, budget-conscious models now address speech and image workloads, respectively. They come at a fraction of the cost (think, 70-80% cheaper than the usual suspects). For startups or teams watching the purse strings, that’s welcome news.

Announcements That Caught the Community’s Ear

No matter how many developer events you’ve attended, there’s always that feeling when the community collectively leans forward in their chairs. Here’s what sent ripples through the crowd at DevDay 2025:

• ChatGPT Apps and Apps SDK: This was perhaps the most tangible shift in how we’ll build and use automations. The integration possibilities feel, dare I say, endless. I chatted with several developers during the follow-up Discord roundtables—everyone seemed to have one “aha!” moment imagining their favourite workflow made conversational.
• AgentKit: Production-Ready Agents Without Tears: Having spent yesterday hacking with the demo, I’m genuinely buoyed by how approachable this tooling has become. Yes, there are still the odd gremlins (is there ever a big release without little quirks?), but AgentKit is likely to open doors for hundreds of teams who’ve previously found AI agent deployment too daunting.
• Sora 2 in the API: The prospect of on-the-fly video generation in client-facing applications immediately resonated. I ran a test workflow—admittedly a quick and dirty one, fresh off Make.com—embedding short video snippets in automated sales follow-ups. Worked a treat, and the client was genuinely dazzled.
• Enterprise-Grade Codex Integrations: As a certified Slack addict, I was thrilled to see native integration with the messaging staple. Now our daily chatter can feed directly into code suggestions, project automations, or documentation updates—hugely helpful for distributed teams like ours.
• Budget-Friendly Mini Models: There’s always a trade-off with model size, but for those automating volume tasks (like sorting vast email queues or auditing images for compliance) these new releases are an absolute lifeline. I’ve already migrated part of our internal workflow and noticed a pleasant dip in hosting costs.

A Word on GPT-5 Pro

Having used every GPT iteration since the early days, GPT-5 Pro struck me as a mature, reliable partner for tricky prompts. It’s noticeably better with region-specific idioms (bless whoever trained it on so many British turns of phrase) and, crucially for us in marketing automations, it delivers more consistent tone continuity across large batches of generated content.

DevDay as Seen by an Attendee

While the flashiest bits naturally come from the product launches, I always make time for the softer moments—networking, hallway chats, and, of course, the occasional fanboy selfie. Getting a glimpse into the minds behind OpenAI underscored for me just how people-focused their work aspires to be.

One highlight was catching Greg Brockman and Sam Altman discussing the social promise of AI. There’s a real sense, from both their keynotes and side conversations, that OpenAI isn’t interested in just dropping shiny tech for the sake of it—they’re staking a claim as social architects in the AI landscape. Not everyone takes to this with enthusiasm, but I do find the vision engaging, especially when you look at the practical, city-level AI projects that were discussed this year.

• San Francisco’s Embrace of AI — Daniel Lurie and Jony Ive added refreshing perspective. Their sessions, peppered with local anecdotes, showcased how public policy and AI can actually meet in the middle—as seen in housing reforms and sustainability projects powered by OpenAI’s tools.
• Everyday Stories — Chatting with fellow builders, I kept hearing fresh stories of AI unblocking bottlenecks from charity work to small-scale entrepreneurship. You can tell the tools are working their way into all sorts of hands, often in unexpected directions.

Thematic Sessions—Building the Future Day by Day

Plenty of folks come to DevDay just for the hands-on learning, and this year’s workshops didn’t disappoint. I juggled between note-taking and frantically bookmarking slides throughout:

• Sam Altman’s Keynote: Love him or loathe him, Altman knows how to provoke. His forward-looking take on broadening access to AI was underscored by concrete examples—think not just what we can automate, but who gets a seat at the table as we do it.
• Context Engineering & Coding Agents (Cursor Team): I was particularly drawn to this workshop. It went beyond dry technical explanation and delved into context engineering—how systems make sense of messy, real-world prompts. This session genuinely changed my design approach, especially when deploying bots at scale on n8n and Make.com.
• Orchestrating Agents at Scale: This session was a sprint through best practices for architecting agent-based solutions, not just for small demos but for enterprise deployment. Some of the tips for managing agent workflows at volume were worth the price of admission alone.
• AI for San Francisco: Tackling public sector problems with modern AI—housing allocation, transport smoothing, social care applications—this panel was brimming with practical examples and honest, sometimes delightfully blunt feedback from city officers.

How to Revisit DevDay: Materials, Recordings, and More

If you (or your team) missed the live action, OpenAI has gone the extra mile to document every panel and talk. Everything is streaming on YouTube and easily navigable through their official playlist. In my view, this is a windfall for those of us running distributed businesses or simply unable to hop on a plane to San Francisco.

• Keynote and Panels – Watch or rewatch the centrepiece ideas at your own pace.
• Breakout Sessions – Hunt for gems within specific technical topics; I’ve saved at least half a dozen for in-depth study later.
• Live Demos – My favourite bits. Nothing beats watching the presenters scramble to fix an unscripted hiccup in real time.
• Vlog Summaries – For anyone needing the “too long, didn’t watch” version, the community vlogs and highlight reels offer speedy overviews.

Industry Reflection: What’s Changed After DevDay?

I won’t sugar-coat it: OpenAI is moving at a canter. Each new set of releases gives our workflows a shake-up, and this year, some changes are already visible even outside of the core AI sector. Here’s what’s resonating around my circles and among industry peers:

• Faster Prototyping: The combination of Apps SDK, AgentKit, and updated Codex means you’re no longer wading through endless proof-of-concept purgatory. Our last two client pilots went from rough outline to sandbox deployment in record time.
• Democratisation of Advanced AI: I know it sounds lofty, but with affordable mini models and cloud deployment patterns, even small teams can now handle tasks reserved for monolithic enterprises only a few years ago.
• AI Beyond IT: Most telling of all is how often I’m hearing from non-technical leads—marketing, HR, logistics, even legal—wanting to embed AI agents in their processes. The barrier is falling, and so is the mystique.

My Personal Take—Trials and Triumphs

I couldn’t resist tinkering with AgentKit the moment it was made available. Within a few hours, I’d spun up a prototype of a sales assistant pulling data from our CRM, prepping daily digests for the team, and triggering tasks over in Make.com. Sure, there were the inevitable teething issues—a typo here, a missing permissions flag there—but nothing insurmountable. The big win? Watching the team pick up the tool without the usual weeks of onboarding. That alone saved us a fortune, both in cash and patience.

There’s still the odd bramble to clear. I bumped into minor bugs when linking AgentKit outputs to our n8n flows, and for one stubborn workflow, a custom integration was necessary. Let’s be honest, it wouldn’t be a real-world launch without a bit of hair-pulling. Yet, as my old computer science supervisor used to say, “no pain, no gain.”

DevDay’s Broader Context: Social Impact and AI Ethics

One message I couldn’t shake off this year was the interconnectedness between technical advances and societal outcomes. There was a noticeable increase in sessions devoted to bias mitigation, responsible AI deployment, and opening pathways for civic uses of AGI.

• AI for Good – Project showcases from non-profits and municipal bodies made it clear that ethical deployment isn’t just a footnote; it’s embedded in real, impactful work.
• Accessibility and Transparency – OpenAI’s new documentation standards and open demo policies are making it far easier for those not already on the inside track to get up to speed.
• Developer Wellbeing – I noticed thoughtful nods to work-life balance and the need for AI to serve actual human priorities, not round-the-clock hustle. A welcome shift, honestly.

How to Make the Most of DevDay Content

I get a bit evangelical about learning from events after the spotlight fades. Here are my top tips for genuinely making OpenAI DevDay content work for you or your organisation:

• Schedule Group Replays – I ran two quick lunch-hour screenings with my team. Pausing to discuss immediately after sparked more ideas than any slide deck could express.
• Bookmark “How-To” Demos – Especially if you’re using Make.com or n8n, some of these sessions are goldmines for jumpstarting production-grade automations.
• Experiment While Watching – Open up your editor or dashboard and shadow the demos live. A bit of hands-on fumbling goes a long way.
• Feed Back Into the Community – Don’t keep your findings (and bugfix hacks) to yourself. A quick post on developer forums or Slack channels can save other folks a mountain of frustration.

Future Trends Spotted at DevDay

I’m always looking for the “next big thing” beneath the surface. Peering through the sessions and side talks, a few trends stood out:

• Conversational Automation Is Here – Embedding human-like chat agents into routine workflows is quickly moving from “nice to have” to “must have”.
• Hybrid Model Stacks – More teams are mixing large and mini models—balancing cost and firepower depending on the task at hand.
• Contextual, Real-Time Analytics – With context-aware agents, real-time business intelligence is taking on a much more nuanced, situation-aware shape.
• AI for All, Not Only Engineers – The way OpenAI’s tools are being reimagined by marketers, designers, educators, and, indeed, city planners is frankly refreshing.

Voices from the Floor: Community Impressions

Here’s a sampling of quips and thoughts I overheard during breaks and panel Q&As—because sometimes nothing beats a bit of raw, boots-on-the-ground feedback:

• “I spun up a reporting workflow during the lunch hour. Didn’t even need to write a line of new code. That, for me, is magic.”
• “Finally, video generation I can plug straight into my site. Been waiting on this for months.”
• “AgentKit feels like the missing puzzle piece. I was hesitant at first, but it’s brilliant for integrating sales outreach on n8n.”
• “It’s great to see accessibility, not just in tech features but pricing structures, too.”
• “Love the focus on civic applications—AI that helps, as well as impresses.”

Closing Thoughts

As I wrap up this year’s reflections, I find myself feeling genuinely optimistic. Whether you’re a serial entrepreneur, a code newbie, or, like me, somewhere in the middle, this year’s OpenAI DevDay delivered more than just technical sizzle. The focus on lowering entry barriers, broadening use-cases, and partnering with civil society turned what could have been “just another” dev conference into something more substantial.

If you haven’t yet, I encourage you—pull up the official playlist, grab a notebook (or your favourite task app), and tune in for yourself. I’ve already seen the benefits ripple through my own practice and clients’ businesses. Here’s hoping you do, too.

And, as we say in the UK, “Give it a whirl!” Even if you stumble a few times, you’re in excellent company.

See you next year at DevDay!