Aaron Grando

@ericvishria @sarthakgh This is more a problem with AI spend getting locked at a previous underestimation than a real issue.

Seeing this all over. The efficiency and rationalization push is coming. Initially it was just getting everyone to use AI all over. "Opus everything!" Now the spend is material and companies will rationalize.

@thenanyu Unironically having a background in Magic: The Gathering has been hugely valuable to my career

People are like "I have a design background" or "I have an eng. background" If I really thought about it— I have a classical music background, and a DotA background. I was 6 years into my dev career before I had logged more hours programming than either of those things. A lot of how I practice teamwork, leadership, getting results, etc. comes from those activities (both done at an unreasonably serious level)

@benjitaylor Benj we need dim mode back, white text on black S U C K S

I’ve been extremely impressed at just how much every engineer at 𝕏/xAI genuinely cares about design and craft. Everyone wants to go the extra mile, cares about the tiny details, and wants to work closely with design to make it happen. Breath of fresh air.

Damage done, for what

@himanshu__sriv @eastdakota It’s incredible that “it’s not this it’s that” is always 2nd to last paragraph, even in minimal two-sentence posts.

@eastdakota The tool didn’t change the job—it changed the leverage. Good engineers now think better, not type faster.

Yup. Feels completely unfair. But we now have electric screwdrivers and if you’re still insisting on only using a manual screwdriver, you’re a dinosaur. PS - not just engineering.

On occasion my sessions with Claude will devolve into guttural impulses, “go” “do” “y” “continu” “keep goin” “undo dat”

@deredleritt3r @NateSilver538 I mean the wings aren’t flapping, any object with a taper to it can glide if you strap a powerful enough engine to it. Doesn’t mean it’s flying.

@NateSilver538 Particularly embarrassing to continue to hold such views after the "flying machines" have already started flying

Since the technology is changing so fast, AI takes really expose who is able to think on the fly versus who is totally impervious to new evidence because they carved out some "brand" years ago.

Also (as of earlier this week) the official Figma MCP/plugin that's available in the Claude desktop app as a connector(?) does not seem to use the generate_figma_design tool that the Figma MCP that CLI Claude Code uses pretty readily. Been stumping my designers who can't swallow working in the terminal. x.com/hugethanks/sta…

@trq212 This is awesome. I REALLY hope you can show the process on a fresh Mac with no dependencies installed, this has been a hurdle as I've been onboarding teammates.

I'll be doing a livestream with Figma on March 31st on how to use Claude Code with the Figma MCP to collaborate between engineers and designers. Hope to see you there, you can sign up below!

The reality lane for companies that aren’t AI labs > team sizes shrink but not because AI takes the work directly, but because AI triggers new process to be implemented that just happens to work better with smaller more agile teams > teams and processes are the units of productivity to care about not individuals or hours. More smaller teams will exist and they will be dynamically process driven; process heavy, but flexible processes. > the same reduction in time to create and deploy software will happen to processes as agents easily abstract and smooth over frequent process iterations for the people involved. > client budgets move with the economy, there is no alpha for applied AI or token resellers (fully captured by model providers) and what we now consider core biz capabilities are recentered, boosted by AI. Smart or at least serviceable use of AI will be assumed, like email. Not a selling point. > the only technology advantages you have are time bound (they always have been but AI speeding progress makes windows of opportunity tighter and more noticeable)

@tomfgoodwin @kylewgrove Because for a long while the only simple control surface for repeatable AI instructions was an agent system prompt. We’re past that now and a well configured single agent with the right tools can access lots of background and know how without the segmentation.

I’m surely being stupid. But if AI is rather unconstrained by expertise or capacity or to some extent speed Why do we need to divide tasks or departments to 9 agents ( the marketing agent, the optimization agent etc ) to each do one thing. And then another agent to manage the swarm. Cant one agent just be doing it all you know. It seems very skeuomorphic. Will we have HR agents to make sure the agent agents are being looked after ? A office canteen manager agent to feed the agents ? Seems daft

@amorriscode @Miles_Brundage x.com/hugethanks/sta…

@Miles_Brundage Will echo other folks who have responded: we are online, we are deeply sorry for the pain we cause customers, and we are working to make it better. I can't make all the issue go away but if you have specific complaints about product I will do my best to amplify when necessary.

I'm a bit worried Anthropic has an org-wide case of AI psychosis that makes them think Claude is good enough that they can ship random product features without breaking things, but in fact they *do* keep breaking things, and they're not online enough to notice people complaining

Not enough said about how the desktop UI app (non-terminal) version of Claude Code seems to perform much worse than the terminal version using the same plugins/connectors/skills/MCPs (naming differences between the two surfaces here are pretty bad) and just how hard that’s making it to onboard nontechnical users. frustrating because it’s SO CLOSE to being great. @claudeai happy to share more, reach out. my advice will be to work on getting a designer set up with the figma mcp in the Claude desktop app to the point which Claude can successfully run the generate_figma_design tool to convert an in browser prototype to an editable figma file. Harder than it sounds.

@ericzakariasson Started with 5.4 x.com/hugethanks/sta…

whats going on with chatgpt these days? almost all responses ends with clickbaity questions "if you want, i can tell you the one mistake that almost everyone forgets"

why the hell can't i awkwardly cmd+v paste an image into my browser bar and search based on the content of the image like i can with my boy cc

Wow I really hate the bait that 5.4 leaves at the end of some responses. “If you really want to know, I can also tell you something that might make tomorrow’s meeting easier: There’s a very simple way to think about X that almost always lands and avoids all the pitfalls I just mentioned. Want to hear about it?”

@michaelmiraflor this graph starts in a year when we all were given thousands of dollars of free money and the economy suffered a once in a lifetime instant shock…how did this look through the 2010s?

Absolutely a leading indicator of white collar crisis.

The management adoption reported here should be offensive to everyone trying to do real work with AI. Have largely experienced this as a midcurve phenomenon. Top of org managers and bottom of org managers are AI forward and moving hyper fast. Mid management consensus is still wait and see (or, ignore and preserve).

Can't stop staring at this image and thinking about what it means for the future

Great team theory is real. When you work with or experience the work of a team at the top of their game the feeling is unmatched. Better than the sum of its parts and sadly often fragile enough that by just removing one person the whole thing changes. Related, OG Verge still lives in my brain, receipt:

You all NEED to stop naming your products after office supplies. SO BORING