Rahul Dave

@emollick Still feel Claude code was best when faced with the 400k context limit, and where it would send sub-context to appropriate sub-agents..now the new Ruling in the harness seems to be confused, and the context way more diffuse

Infinite context windows seem to present a very large problem to using AI. Today's models already leak too much old information into current responses, a distraction that is part of why they are cognitively exhausting to use I don't want to work with Borges's Funes the Memorious

@dbreunig We should expect this, shouldn't we, because most writing, especially engineering writing with jargon and all that lacks clarity. Bad writing is in-distribution, perhaps?

We need a name for this, because Armin is putting his finger on a problem that’s everywhere: people running their writing through an LLM because they think it makes it clearer, when in actuality it sands off all the detail.

Most people don’t understand that the Indian bigco CEO phenomenon is mostly an immigration story. Indians are FAR more likely to be tied to employment visas than other nationalities, so they couldn’t easily start companies. The 1990 Immigration Act created the modern H-1B/EB green card system. As Indian demand exploded, the 7% per-country cap turned into decades-long backlogs for Indians. Indian tech workers stayed tethered to sponsoring employers in a way Europeans, Russians, Taiwanese never had to. Founding a company means risking your status, resetting a green card path, or finding another workaround (now usually O-1 or EB-1). European/ Russian /Taiwanese immigrants don’t face the same trap. Their countries don’t hit the 7% per-country cap, so demand stays under the limit. A German or Russian engineer on H-1B can get a green card in 1-2 years and leave to found a company. An Indian engineer doing the same job has to wait 20+ years. The 1965-1989 Indian cohort was much smaller but not yet trapped by today’s H-1B lottery and India backlog machine. That’s why you see so many Indian founders from that era: Vinod Khosla (Sun Microsystems), Sanjay Mehrotra (SanDisk, before becoming CEO of Micron), Kanwal Rekhi (Excelan), Suhas Patil (Cirrus Logic), Desh Deshpande (Sycamore Networks), Pradeep Sindhu (Juniper Networks), etc. My dad is a 1972 IIT grad who came to America for a PhD. Most of his IIT friends are successful entrepreneurs. My cousin took the same exact path (IIT>CMU) in the 1990s and most of his friends worked their way up corporate jobs because they needed employment sponsorship. IMO this is bad for America. We took the highest-conviction risk-takers on earth, people who crossed an ocean and left their families behind, and forced them into the lowest-risk career path. Fortunately this has been loosened in the 2010s with O-1 and EB-1A workarounds but it’s still much more challenging for Indian or Chinese founders.

1/ cricket-lens.sfunicorns.ai/chat/ Cricket Lens is now open to use by everyone. Try it out and let us know what you think - would love feedback!

racism on the TL today in full force. reminder to my caramel kind to just continue being excellent and winning. the neanderthals will die and no one will remember them, that is their nature and destiny. yours is to love, persevere, and excel.

First they came for undocumenteds, you said nothing because you’re there legally. Then they issued visa bans on multiple countries and set ICE on rampage. You stayed silent because it didn’t affect you. Now they’re targeting those already in the legal green card pipeline: forcing H-1B and O-1 holders to leave the U.S. for consular processing. Don’t speak up. Green card holders are next. Then naturalized citizens. Then US born children of non-citizens. Then …

@wesbos @appfactory Will say there is something perverse about agents doing commits and docs and all that: is developers surviving because we write badly. AI popularizes the median and the median is bad, writing or coding, and now we see so many docs which are just slop...

@rahuldave @appfactory yeah exactly, I want both! And I think WebMCP is that

Give me your hottest takes on generative UI, MCP apps, agents returning widgets, and chat as the final interface Working on a talk for JS Nation and I want to spotlight some differing opinions

Will be playing for sure! Was trying to solve for lineage from agent turns to source changes and fork abilities as a result of that. But if you think about it agent logs and source changes don't easily belong in the same repo, unless you are willing to tolerate bloat. So you need to start from a deeper primitive like you do to hang multiple storages together: almost like lakehouse formats plus version control...

@rahuldave oh cool, I think a lot of people are circling around this direction. would love to see what you think if you get a chance to play around :)

babyagi has ~200 citations, but 0 papers... i just published my first paper on arXiv 😆 "The Log is the Agent: Event-Sourced Reactive Graphs for Auditable, Forkable Agentic Systems" arxiv.org/abs/2605.21997 the case for agents that coordinate through persistent replayable state — no conversation loops, no workflows, no A2A — with auditability, forking, and causal lineage built in. check it out and let me know what you think!

@wesbos @appfactory It's the size of the interaction no? Like ai is better than me at writing commit messages now...and at finding all your stuff online or twitter, but replying to this is faster than having ai do it. So latency issue...

@appfactory but like, even this interaction, you would have rather typed "Show me the latest posts", then "please reply to this post with..." instead of clicking "reply". I want to like this, so I type "like this post"?

@HedgieMarkets What about strong distillation to weaker models, or the use of weaker frontier models in agents? Task orientation is not a bad thing?

🦔Microsoft canceled its internal Claude Code licenses this week after token-based billing made the cost untenable, even for a company with effectively infinite cloud resources. Uber's CTO sent an internal memo warning the company burned through its entire 2026 AI budget in just four months. American AI software prices have jumped 20% to 37%, and GitHub (owned by Microsoft) is dropping flat-rate plans for usage-based billing across its products. My Take The AI subsidy era is ending in real time. The same company that put $13 billion into OpenAI and built the Azure infrastructure powering most of Anthropic's compute just looked at the bill from a competitor's coding tool and decided it was not worth paying. That is not a productivity failure on Anthropic's end. Token-based pricing is forcing every enterprise customer to confront the actual cost of running these models at scale, and the number turns out to be far higher than the flat-rate experiments suggested. This ties directly to my Gemini Flash post yesterday. Anthropic, OpenAI, and Google all raised effective prices in the last six months. Enterprises that built workflows assuming AI costs would keep falling are now watching annual budgets evaporate in months. Two outcomes look likely from here. Either enterprises scale back AI usage to fit budgets, which slows the revenue ramp the labs need to justify their valuations ahead of IPOs, or the labs cut prices and absorb the losses, which makes the unit economics worse at exactly the wrong moment. Both paths land in the same place, the numbers stop working, and somebody has to take the writedown. Hedgie🤗

@nbaschez It is a lovely piece of software! Really like the idea and some directions you are thinking of including putting annos in yaml ( maybe in frontmatter? ). Thinking it can be used for multiple agents to communicate as well: needs an additional annotator metadata...

Main thing I learned yesterday is I have a lot of good friends on here Thank y'all so much for the support ❤️🙏 Ok now I'm gonna go ship a whole bunch of little polish to this thing

Hilarious

I got board of opening spreadsheets with google sheets. So I've built `spread` a native spreadsheet viewer. Built with @zeddotdev's GPUI. Can open xlsx, csv and parquet. Will open a 30M row parquet file in <100ms. Renders most xlsx formatting. github.com/samuelcolvin/s…

@vboykis Ditto. And worse i feel bad if i am not talking to the agents on a weekend...

Something happens to my brain after agentic coding that I can’t describe. It’s like cognitive offloading which folks have already written about, but even more. It feels like I can’t think through problems anymore. Like a fog. Using agentic but losing my hard-won agency.

@DimitrisPapail The architectural boat tour, and just about any architectural tour of chicago is a lot of fun!

Chicago is stunning. Perhaps the most beautiful of American cities, and a real one at that.

1/ We opened Cricket Lens to early access this week. A sampling of queries from IPL franchise analysts, analytics firms, and independent professionals: A sampling of queries 👇

New blog post. Just thoughts this time. maxliani.wordpress.com/2026/05/11/pas…