Buffer Engineers

Buffer's API moved into general availability yesterday! 🎉 One API, eleven social platforms, official partnerships with: Meta, LinkedIn, X, and others, and the first time Buffer's full surface area is open to builders. In the video, our Staff Product Manager Amanda Marochko walks you through where to start: - The developer portal, - the GraphQL explorer, - and what our MCP server unlocks if you want an AI agent publishing on your behalf. The work behind this came from engineers, designers, product, support, marketing, and the partner teams at the platforms that matter most. What this opens up in practice: - Creators layering Buffer into a Slack bot or a Lovable app, - marketing teams plugging Buffer into the AI workflows they already run, - and developers and agencies building full integrations on one schema with no SDK required. Most of what's live today sits in publishing, scheduling, and content creation. Analytics, community, and the rest of Buffer's surface area open up next! The shape of that roadmap will be set by what builders push us toward. If you ship something on the Buffer API, send it our way. We'd love to see it! ⬇️ 💬

I've been experimenting with using different AI models for different parts of my workflow, and I've landed on a setup that's been working surprisingly well. For a while I assumed the goal was to find the single best model and route everything through it. One tool, one workflow, less to think about. But the more I actually worked across different kinds of tasks, the more I noticed that "best" depends a lot on what you're actually doing. When it comes to backend work and the bigger architectural decisions, Codex has been working well. It's good at the kind of thinking you do before you write any code, like reasoning about how services should talk to each other, where state should live, what the tradeoffs are between a few different approaches. It doesn't just answer the question I asked, it tends to catch the thing I should have asked about. For anything where the structure matters more than the syntax, it's been solid. Claude has been the opposite end of that for me. Frontend and design work is where it consistently does better. When I'm building out UI, working through component structure, or trying to get something to actually look and feel right, the output tends to be cleaner and needs a lot less back and forth to get usable. It seems to have a better sense of the small details that make an interface feel considered rather than just functional. I'd guess it'll keep shifting as these models change. But for now, matching the model to the type of work has saved me real time. Curious how others are approaching this. Are you splitting work across models or have you found a single one that holds up well across everything? Always interested to hear what's working for people. #ai #claude #codex #anthropic #openai

Are you an engineer who always defaults to action? This will definitely help your career, but watch out because it doesn't always work in your favor! Let me tell you a story. I'm a big believer in defaulting to action and have a great sense of ownership. I get this urge to always surprise everyone positively with the amount of things I'm able to do in short periods of time, especially that as an engineer I've got means to do so. Threads (the social network) implemented topic fields via API? No problem, I'll put it together within 2-3 hours and have it ready for release at Buffer, customers will be thrilled! They were, but at the same time, when we were reflecting on our sprint as a team, it came out that my colleagues felt the pressure to adjust their workload and add this as an extra work to their plates - specifically mobile apps engineers and designer. After all, we want product parity and excellent UX for anything we release. It was a great feedback for me. It is when I’ve realized that my style of defaulting to action is not always positive. I need way better communication that what I do is mostly showcasing and sort of MVP, and not 100% ready for release product that requires us to drop everything else (especially in design and apps) and focus on that new shiny feature instead. So that’s my learning to be more cautious about communication, prioritization, and jumping on features like that in a smart way, with a plan prepared, not hooray-style. Always remember, working in tech is a team game, and nearly each action you take impacts people in this way or another, even when you don't expect it! Clear communication is clear to avoid situations like this. The situation I described - it still could have been done in the same amount of time, equally good, and with less stress, if only I communicated with my peers better. If there's no communication and alignment - it's not worth it!

Linear just shipped a feature called Diffs that sounds pretty interesting. The idea is simple: review pull requests without leaving Linear. Read diffs, leave inline comments, approve or request changes, and merge, all from the same tool where your issues and projects already live. If you've used Linear with GitHub, you know the current integration already links PRs to issues. Diffs takes that a step further by pulling the actual code changes into Linear so you can review them in place. It's not trying to be a full GitHub replacement. Your CI, checks, and existing workflows all stay the same. It's more like a focused lens into the review process from inside the tool where your planning already happens. What stood out to me from the docs: - A new "Reviews" sidebar section that separates PRs needing your attention from ones you authored - Unified and split diff views with a keyboard shortcut to toggle between them (⌘+B) - Inline comment threads that sync bidirectionally with GitHub. Reply in either place and it shows up in both - Notification controls that let you filter out bot noise, which is a small thing that probably matters more than it sounds - Keyboard shortcuts to navigate quickly (G+R to jump to reviews, O+R to open a specific one) The bidirectional sync is the part I'm most interested in. One of the friction points with any "review code outside of GitHub" approach is that it can fragment conversations. If comments actually stay in sync across both tools, that removes the main reason not to try it. Setup looks straightforward. A GitHub org owner updates the Linear integration to allow code access, selects which repos to include, and then individual team members reconnect their personal GitHub accounts. Existing PR links and history in Linear are preserved. There are some limitations worth noting. Per-commit reviews aren't supported yet, so reviews happen at the PR level. Draft reviews started in GitHub don't sync into Linear. And rich CI annotations (like inline failure locations) aren't displayed yet either. It's clearly early, and Linear says the experience will keep evolving based on feedback. If anyone's already using this, curious to hear how it's working in practice. linear.app/docs/diffs #Engineering #Linear

A secret hack to actually get your request for feedback answered? Start with some self-feedback to make it easier for the other person (e.g., your manager) to just "add to the list" instead of figuring it out whole. Example: You want to get some feedback on the project you led recently. Instead of asking your manager "any feedback about that project I led?", you can do it even better if, instead, you go like "In the last project I led, I know I wasn't updating the stakeholders regularly, possibly leading to some lost trust. Did you notice any other things I could improve in my project management skills?" It requires some amount of self-reflection and work from your side, but that's actually a trait you should already have with a growth mindset!