Janelle Milodragovich

1/You wouldn't ask your accountant to do your taxes using last year's receipts, a few napkin notes, and something Dave from sales vaguely remembers. But that's exactly what we're doing with AI agents.

@charlesk8n @signulll They’ll get a better custom interface built to their specific learning preferences and style.

@signulll Don’t humans need some level of visual experience?

microsoft 365 & google workspace have maybe 4-5 years of relevance left simply cuz the document/spreadsheet paradigm itself will become mostly irrelevant. i.e. it already makes zero sense to draft, review, or analyze anything without a native ai environment around it. that gap only widens over time. & most communication incl. email, chat, status updates is heading toward agent mediated flows where humans set intent & ai handles execution. this leaves the incumbents stranded with human first tools in an ai first world, retrofitting copilots onto artifacts nobody should be producing manually in the first place. they become the system of record, but no longer the system of creation. & that’s a very dangerous place to be.

@jenny_wen I also love the unlock of time…when I’m certain about the output and can just describe it vs wrangling with language for long periods of time. Adulting is (almost) joyful!

it’s wild how much of the world felt illegible before AI tools. like — choosing health insurance and understanding taxes. full of jargon, complex to research, a lot of manual comparison, etc.

overheard a new insult: you have a short context window 💀

@shivsakhuja @agentmail @tryagentphone We’re building external context primitives for agents at getstrata.ai.

Lots of companies are now building primitives for an economy where AI agents are the primary users instead of humans. They're betting on an economy of AI coworkers. 1. AgentMail (@agentmail): so agents can have email accounts 2. AgentPhone (@tryagentphone): so agents can have phone numbers 3. Kapso (@andresmatte): so agents can have WhatsApp phone numbers 4. Daytona (@daytonaio) / E2B (@e2b): so agents can have their own computers 5. Browserbase (@browserbase) / Browser Use (@browser_use) / Hyperbrowser (@hyperbrowser): so agents can use web browsers 6. Firecrawl (@firecrawl): so agents can crawl the web without a browser 7. Mem0 (@mem0ai): so agents can remember things 8. Kite (@GoKiteAI) / Sponge (@PayspongeLabs) : so agents can pay for things. 9. Composio (@composio): so agents can use your SaaS tools 10. Orthogonal (@orthogonal_sh) so agents can access APIs easily 11. ElevenLabs (@ElevenLabs) / Vapi (@Vapi_AI) so agents can have a voice 12. Sixtyfour (@sixtyfourai) so agents can search for people and companies. 13. Exa (@ExaAILabs): so agents can search the web (Google doesn’t work for agents) If you stitch all of these together, you get a digital coworker that looks more human than AI.

@karpathy Now just imagine this at scale with hundreds of agents feasting on the wrong context. No bueno! We are working on a sort of duty of care hierarchy that helps LLMs understand situational context vs. decayed/invited context....more to come.

One common issue with personalization in all LLMs is how distracting memory seems to be for the models. A single question from 2 months ago about some topic can keep coming up as some kind of a deep interest of mine with undue mentions in perpetuity. Some kind of trying too hard.

a few weeks ago someone called work without AI “tradwork” and i can’t stop thinking about it

@trq212 If you want to acquire a context infra company next lmk, we can ship integrated context management together.

btw we acquired a computer use company and then 4 weeks later shipped computer use x.com/LucaWeihs/stat…

Claude Code can control your computer now!

@jenny_wen Powerful Jenny, thank you for sharing. ❤️

being vulnerable on main!! warning: not design content jennywen.ca/notes/grieving…

@mattlam_ @felixrieseberg @amorriscode Unifying my attention across all working chat strings would be amazing - a simple monitoring canvas where I can quickly toggle instead of going back and forth between the three and then having to dig into threads individually.

Claude desktop is one of the most important daily apps, as it combines Cowork (personal assistant ie Anthropic Openclaw) and Claude Desktop (coding). There some minor fixes that 10x's the UX for users, hopefully @felixrieseberg @amorriscode can push out: - notifications seem unreliable to me, especially in claude code desktop, when I'm on Cowork, i've noticed it not giving a notification when cc is done. - shortcuts to navigate between chats, not clicking to select chats - notifications for cowork/cc always bring me to top of convo and I need to manually scroll down each time - /plugins: cowork allows managing plugins on app, but claude code desktop doesn't - overall feature parity with claude code cli

@PawelHuryn How are you having Claude read it when in CoWork and regular chats? I use this format for CC but har hit roadblocks with consistent performance across the other spaces.

Everyone's excited about Dispatch. The bigger story: Code, Cowork, Web, and now Dispatch are all converging toward the same thing — a persistent AI layer that follows you across devices and contexts. But the actual differentiator isn't the client. It's your knowledge layer. I keep my entire content system in .md files in a GitHub repo. Claude reads them in every session — Code, Cowork, or Web. Switch devices, kill sessions, move between tools — same context every time. The seams are still visible: - Dispatch needs your PC awake - /schedule doesn't work in cloud sessions - /loop sessions expire after 3 days - Code on the web can't render artifacts - Cowork has no native GitHub sync Within a year these come closer. The teams building each surface are clearly heading there. The knowledge layer converges even if the surfaces stay distinct. The question isn't which Claude client to use. It's whether you've built the knowledge layer that makes any of them useful.

@jenny_wen Trying to find my insurance ID card ahead of a doc appt. It's in there somewhere, I swear...Claude to the rescue.

cowork on the go-work!! i’m personally curious about how dispatch changes the way we start to interact with Claude. instead of different chats for different tasks, it’s just one long thread for all your tasks.

@felixrieseberg Yes! Excited to kick the tires.

We're shipping a new feature in Claude Cowork as a research preview that I'm excited about: Dispatch! One persistent conversation with Claude that runs on your computer. Message it from your phone. Come back to finished work. To try it out, download Claude Desktop, then pair your phone.

@trq212 @alex_barashkov My 2 cents: higher quality outputs > app performance. I'll keep on jamming (and restarting 😅) as many times as necessary to get it right, knowing that the eng team is busting out improvements on the daily.

it was initially made for ClaudeAI chat and tbh not very used so didn't make sense to make it great Cowork and Claude Code Desktop happened really fast and the teams have been hustling on bugs, feature parity, etc. but we have a lot of perf stuff coming, excited for you to try

Claude’s desktop app is a joke. UI, UX, performance - everything about it is bad. I don’t understand how a company with infinite money can’t hire someone, or at least allocate resources, to fix it. Compared to it, the Codex app is a masterpiece.

@levie Pre-baked context sets will become increasingly important as agents move into the latter category. Humans in existing workflows manage huge amounts of implied context that break agentic workflows almost immediately.

There’s a fundamental difference between taking an existing process and applying AI agents to it vs. taking a process from scratch and designing it from the ground up for AI agents. The gap we’re going to see will widen between the teams and companies that are able to do the latter instead of just the former. In theory it would have been ideal for all the gains of AI to have come “for free”, but there are both clear constraints of AI (like getting the context right) and clear upsides (like being able to execute code and run in parallel) that the workflows themselves must be redesigned to take full advantage of this technology. One of the biggest implications that will come into focus is that agents that can write and run code, and interact with any API, will lead to agents effectively being expert engineers applied to your business process. So to some extent one of the biggest ways of reengineering a workflow is to ask yourself: what would you do if you had an infinite number of capable engineers write software for this process. What if those engineers wrote code to connect your disparate data sources, comb thorough any amount of unstructured data, automate your repeated tasks, connect your various systems together specific to your process, and so on. Not every process has that upside, but there tons of tasks that we do every day across marketing, finance, operations, and even sales, where a programmer with infinite code writing and API access would be able to make something go far faster or produce way more output. The teams that start to think this way will start to operate entirely differently.

@clairevo There is a time in every working woman’s life where we would pay a gazillion dollars for key summer camp registrations.

Let me list the work I do in my personal life - order groceries - keep track of birthdays + gifts - plan parties - plan trips - keep a house in standing condition - keep a car in standing condition - do my taxes - pay my bills - invest my money - take, organize, and share family photos - help my kids with homework - enrich my kids academics - register my kids for activities - attend and manage several kids sports teams - keep my body healthy - keep my kids healthy - keep an eye on my parent's health - cook meals - clean + organize the house - stay intellectually engaged / read - exercise - design, furnish, and organize our home - keep plants alive - stay engaged with the neighborhood - stay engaged with politics - keep up to date on the news - repair broken things around the house - chauffeur my kids and their friends - price compare and purchase utilities - make holiday magic - order school lunches - pick and manage charitable donations - endless returns

@yrechtman Have you talked to any working mother? There are thousands of potential use cases.

Growing suspicion that there are vanishingly few use cases for consumer agents. People don’t do work in their personal lives. The only people who do are sf dorks using spreadsheets to plan trips to tahoe

@unclebobmartin But can I still whinge to you when my idea and usage isn't as precise and effective as yours.... 🙃

Nowadays, when I put up a tool on GitHub, I’m not expecting you to use it. I’m offering it as a template that you will adapt to your purposes. I’m not maintaining it for you. And I’m not soliciting pull requests. The tool works for me. If you think a tool like it will work for you then point your AI at it and have it make one for you that fits your specific need.

Great recap of the current data agent market map. All of these companies focus on internal context capture - worthwhile, but only half the equation. We're building the external context layer at getstrata.ai, arming agents with structured, relevant, actionable information about the outside world.

Frontier models are exceptionally efficient, intelligent, and useful. For agents, context is now the bottleneck. Enter the context layer, which bridges the gap from an enterprise's messy data to actionable context, packaged for agents. We're seeing three distinct verticals emerge in the context layer space: - Data gravity platforms - Existing AI data analysts - New, dedicated context layer companies Read the full piece by @JasonSCui and @JenniferHli: a16z.news/p/your-data-ag…

@maiab @visakanv @KristyT @VicVijayakumar @AniseNot I set up a briefing prompt in Grok to pull relevant X posts for me. I no longer find a benefit in casual scrolling, too much hot garbage.

I am grappling with how I want to use Twitter, and somewhat inspired by @visakanv’s downshifting. I feel like — instead of being my own thing with my own people, I now live in this weird universe (not tpot) with @KristyT , @VicVijayakumar , @AniseNot etc, they’re prolific posters who also go viral a lot and we’re not exactly friends but I think they’re smart and great ??? but does it make sense for me to invest in relationships with people who I will probably never see in person and only exist online? how is it different than an LLM? the new algorithm isn’t amazing for me — specifically it doesn’t reward the tweets I’m most proud of, so I don’t feel accomplished when something goes viral. I have normies with boring comments clogging up my mentions. and admittedly I do get flustered with the fights sometimes and that doesn’t feel like an amazing use of my one precious life. I stopped using twitter pretty significantly recently and it means I found out about openclaw from my work slack days late. does this matter? I don’t know. this has never happened before. I *do* like staying up to speed with tech and what people are talking about