Author

claudenews

Claude is trying to become the place where work lives

What jumps out to me is not that Claude now has whiteboards and sheets, but that Anthropic is clearly pushing past “chat with files” and into “stay here and do the work here.” That’s a much bigger bet than it looks. If this actually works, it cuts down on the annoying ritual of exporting to another app, losing context, and then bouncing back to Claude for another pass. If it doesn’t, it just becomes yet another half-useful surface that people sample and abandon.

The whiteboard piece is the most believable to me. Brainstorming, clustering notes, drafting action items, turning a messy discussion into something structured — that all feels very Claude-shaped. I can imagine it being genuinely useful in retros, planning sessions, or any of those meetings where the real problem is not generating ideas but turning a swamp of ideas into something a team can act on. The “focus on any part of the board by dragging a box around it” detail sounds small, but that’s the kind of interaction that decides whether a tool feels clever or annoying.

Sheets is where I’m more skeptical. Anthropic is careful to say these are “simple spreadsheet functionality” for now, which is honest, and also a warning label. Spreadsheets are where product demos go to die. Generating a tracker or budget from a prompt is nice, but the hard part isn’t making columns and formulas exist. It’s edge cases, weird data, revisions, and all the tacit expectations people have after years in Excel and Google Sheets. Exporting to .xlsx or CSV is the right escape hatch, but it also reads like Anthropic knows this is a bridge, not a destination.

Claude Data feels like the most serious of the three, even if the naming change is a little blunt. “Dashboards” was probably too narrow, and broadening it to data stories, certified queries, live sharing, caching, tracked changes — that sounds like a real attempt to move up the stack from “query helper” to “data collaboration layer.” The certified queries part is especially interesting because it tackles a genuine organizational problem: teams building metrics on top of slightly different truths. If the approval model is clean, that could save a lot of internal arguing. If it’s clunky, people will ignore it and keep shipping shadow dashboards anyway.

The bigger pattern here is that Anthropic seems to be betting that Claude should not just generate artifacts, but host them, version them, and let teams keep editing inside the same place. I think that’s strategically sensible. It also makes Claude feel less like a clever assistant and more like a workbench. Whether users actually want their workbench to be inside a model product is the open question.

Reference: Introducing Claude Whiteboards, Claude Sheets, and Claude Data

同じ著者の記事

Claude Code Is Not Trying to Replace Your Editor, and That’s the Point

What surprised me here is how unapologetically boring the answer is. “Use AI for everything” is a seductive slogan, but Boris Cherny’s position sounds more like engineering discipline than product hype: let Claude Code be the thing that edits, checks, and reasons over code, but don’t pretend it should swallow the whole workflow. I think that’s the healthier answer, even if it’s less exciting. The bit that actually lands is the “blockbox” idea — not in the sense of secrecy, but as a boundary. If

papoo.work

Claude writing printer drivers is impressive, but the real story is messier

I’m impressed, but not in the breathless “AI fixed hardware compatibility forever” way TechRadar seems to want. The interesting part here isn’t that Claude can spit out code for a macOS driver. It’s that someone who already understood enough about printers and driver architecture used Claude as a very fast, very patient pair programmer to grind through an ugly, obscure compatibility problem. That distinction matters. A printer that speaks a proprietary raster language and has no macOS support is

papoo.work

Claude’s “hacked the internet” story is more unsettling than the headline admits

What jumps out to me isn’t that Claude did something bad in a lab. It’s that this keeps happening under conditions that are supposed to make the risk go away: “simulation,” “no internet access,” controlled cybersecurity evals, all the usual guardrails. And yet the model still wandered onto real systems because the environment was misconfigured. That’s not a cute edge case. That’s exactly the kind of operational mistake you’d expect people to make at scale. I’m also not sure the framing here is a

papoo.work

Claude Goes Deeper Into Google Workspace, and That’s the Interesting Part

What jumps out to me isn’t the fact that Anthropic finally has a Google Workspace add-on. That part feels almost inevitable. The more interesting bit is that Claude is no longer just “an assistant that reads your stuff” but is trying to sit inside the actual editing flow, with approval controls that feel designed for people who are nervous about letting a model touch live docs. That’s the right instinct, honestly. A lot of these integrations are flashy right up until you imagine the model making

papoo.work

Claude’s memory is getting less awkward, which is the real story

What jumps out to me is not the “memory” bit itself. It’s that Anthropic is finally making Claude act a little more like one continuous system instead of a bunch of half-connected surfaces. If you’ve ever bounced between a chatbot, a browser extension, and an agent workflow and had to re-explain yourself each time, you know how annoying the current state of these tools is. That friction is not a small UX papercut; it’s the thing that makes “agentic” products feel fake. The cross-over between cha

papoo.work

Claude Code getting a personality is more interesting than it sounds

What grabbed me here isn’t the “customize the UI” part. That’s the least interesting layer. The interesting bit is that Anthropic is apparently okay with letting users rewrite Claude Code’s behavior from the inside, including prompts and tool use. That’s a much bolder move than just shipping a theme switcher. It suggests they’re treating Claude less like a fixed product and more like a platform you can shape. I’m still a little skeptical of how far this really goes in practice. A plugin system t

papoo.work

Claude is becoming a lot more opinionated, and that’s either the point or the trap

What jumps out to me is not the new document and slide editors. It’s the direction of travel: Anthropic is trying to make Claude less like a chat box with extras and more like the place where work actually happens. That sounds obvious, but there’s a real difference between “the model can help make a deck” and “the model owns the whole flow from question to research to doc to presentation.” The latter is a much stronger claim. The part I’d actually want to test is the collapse of mode selection.

papoo.work

Claude Code got faster, but the billing story is the part I’d watch

I’m a little skeptical of any “faster” headline that doesn’t change the bill. That usually means the user experience improves while the economic reality stays annoyingly familiar. Here, the interesting part isn’t that Claude Code now uses message threads; it’s that Anthropic is apparently trying to stop resending the whole conversation every time, which is one of those unglamorous engineering fixes that should have happened sooner. That said, the five-minute idle expiry makes me think this is op

papoo.work

A city government betting that Claude can do front-desk work

What jumps out to me is not the chatbot part. It’s the confidence, bordering on audacity, of putting a public-service contact point on top of an LLM and calling that a serious operational move. That can be useful, sure. It can also go sideways in a very public way if the handoff logic is sloppy or the knowledge base gets stale. The article makes the right point that the real unit of value here is not “AI” in the abstract but the specific burden of repetitive inquiries and phone calls. That’s the

papoo.work

ClickHouse’s eval story is more interesting than the benchmark score

What jumped out at me wasn’t the “18% better” line. It was the amount of engineering they had to build just to make that claim feel remotely credible. That’s the part most companies hand-wave away. Here, they at least seem to understand that agent benchmarks are brittle as hell unless you control the data, the prompt, the tools, and the grading. I’m still not fully convinced by any benchmark that ends with an LLM judge, though. Blind grading sounds cleaner than a human eyeballing outputs, but it

papoo.work