A good coding agent in a loop, a good file system on the cloud and a good communication substrate can automate everything for you.
Stay tuned for the latter two π
Introducing Claude Science, a new app designed with every stage of research in mind.
Artifacts traced to their code, environments managed on demand, and 60+ optional scientific databases that you can connect.
Available now in beta.
My entire AI stack is now Chinese π¨π³
87% cheaper. same revenue
swaps by task:
1. reasoning / backend brain
Opus 4.8 β Kimi K2.7
benchmark gap: ~8% Β· price: ~11x cheaper
2. code generation
GPT-5.5 β Qwen 3.7 Max
benchmark gap: ~18% Β· price: ~7x cheaper
3. agent loops + tool calling
Sonnet 4.7 β GLM 5.2
benchmark gap: ~3% Β· price: ~5x cheaper on input
4. cheap volume / bulk processing
GPT-5.5 mini β MiMo V2.5
benchmark gap: ~6% Β· price: ~12x cheaper
5. image generation
GPT-Image-2 β Wan 2.5
benchmark gap: ~5% Β· price: ~8x cheaper
6. video generation
Sora 2 β Kling 3.0
benchmark gap: roughly equal Β· price: ~6x cheaper
[ result after 30 days: ]
operating costs dropped 87%, output quality dropped 4% on average, revenue unchanged
the most important that these models will be not banned in a month and i can run them locally
nobody will steal my data and i can learn them as i need
full article drops tomorrow with:
> exact routing logic per task type
> the 2 cases where I still pay for American
> the migration playbook anyone can copy in a weekend
VERY IMPORTANT to get migrated now, while it's not too late
Introducing Claude Tag, a new way for teams to work with Claude.
In Slack, Claude joins as a team member with access to the channels and tools you choose. Tag Claude in and delegate tasks to it while you focus on other work.
Introducing the Open Knowledge Format (OKF), an open specification that formalizes the LLM-wiki pattern into a portable, interoperable format.
AI is only as smart as the context we give it. As we build more advanced, agentic AI systems, they need accurate metadata and context to be useful. But in most organizations, that context is locked inside fragmented data catalogs, isolated wikis, scattered code comments, or the minds of senior engineers. Every time a new AI agent is built, teams are forced to solve the exact same context-assembly problem from scratch.
To solve this, we've announced OKF, a vendor-neutral, open specification that formalizes the "LLM-wiki pattern" into a portable, interoperable format. It provides a standardized way to represent the enterprise knowledge that modern AI systems rely on.
β Just markdown: readable in any editor, renderable on GitHub, indexable by any search tool
β Just files: shippable as a tarball, hostable in any git repo, mountable on any filesystem
β Just YAML frontmatter: for the small set of structured fields that need to be queryable: type, title, description, resource, tags, and timestamp
Weβve also shipped reference implementations to help you hit the ground running, including an enrichment agent for BigQuery, a static HTML visualizer, and live sample bundles on @github β https://t.co/ilhAMCrcTc
β Knowledge Catalog can now natively ingest OKF!
Stop reinventing data models and building bespoke integrations for every new AI tool. Here's more about how OKF works β https://t.co/FR4kJRsgEH
The new moat in the agent era is being the tool agents reach for.
A coding agent doesnβt reinvent a database. It wires up Supabase.
The best devtools companies will make themselves obvious to agents: easy to find, easy to reason about, easy to wire up.
Devtools are entering a golden age, but only for companies that realize theyβre selling to agents now, not just humans.
I love this morning briefing feature!!
Curious if thereβs any plan to support AI meeting note taker? Invite Town as my meeting assistant. Iβve tried Granola, Otter, Fireflies, Read AI, Sentra, Clara and honestly I feel they can just be a feature in Town. @jgreze
@joshzjs Thanks a ton Josh. We're not too precious about the tools or context for now, but you're right we could obfuscate a bit more.
(most of the magic is in the graph of data behind the context)
Thanks for the feedback though.
The next generation ads system should be built around agents.
Now, whatβs the best AI startup for AEO / AI SEO out there?
I want to use your product! Happy to be a pilot user even if you are at MVP stage.
Goal: I want my product to show up in the top 5 list in the LLM answer.
Just tried it - much better product experience compared to Claude / Codex (you can feel it from the onboarding steps alone).
One quick feedback is that the agent can easily tell me internal implementation details such as tool schemas, acc ID, session ID, etc. @jgreze
Weβre thrilled to lead Townβs $55M Series A.
Town is a personal AI assistant that works across the tools you already use - email, calendar, Slack, docs, WhatsApp, desktop, web. It learns how you work and starts proactively pitching in.
People are already leaning on Town for the kind of work thatβs personal and operationally messy: running recruiting pipelines, juggling school logistics, processing handwritten grant requests, prepping summaries, drafting follow-ups, catching the stuff that would otherwise slip.
The longer you use it, the more it picks up: your voice, your relationships, your preferences, your routines, what you actually care about.
Jean-Denis Greze and Tony Vincent are the right team for something this hard. JDG was CTO at Plaid and an engineering leader at Dropbox. Tony led product and AI at Google and design at Dropbox.
Welcome, JDG, Tony, and the Town team, to the a16z family.I
By @arampell and @venturetwins
A lot of the vertical agent space (non-niche) and pretty much all agent builder platforms will be eaten by OpenAI and Anthropic alive.
Traditional SaaS will be eaten by AI-native SaaS alive.
Infrastructure is always a good direction because thereβs always new demand.