Dan Shipper on Claude 4, Every's $2M 'SIF seed' round, and why Anthropic dominates coding

Summary

Dan Shipper, co-founder and CEO of Every, joined from Anthropic's Code with Claude event to discuss the Claude 4 launch, Every's new funding structure, and where Anthropic sits in the developer tooling race.

Every's $2M 'SIF seed'

Every announced a $2 million raise led by Reid Hoffman, with Starting Line VC also participating. The structure is unconventional: investors committed up to $2 million, but Every can draw it down as needed rather than taking the full amount upfront. Shipper describes it tongue-in-cheek as a "SIF seed." The logic is deliberate — Every wants capital to experiment without locking into a specific growth trajectory. The company previously raised $700K in 2020.

Every operates as both a media company and a startup studio, bundling a daily AI newsletter, long-form essays, and software products sold to the same audience. Most of its incubations are separate LLCs. It has already spun out Lex, a writing tool now run by co-founder Nathan as a separate company with seed funding from True Ventures. Shipper is explicit that he wants the writing business to remain the core, with software built around it rather than displacing it.

Claude 4: strong on agentic coding, mixed elsewhere

Shipper got hands-on access to Claude Opus 4 and Sonnet 4 before launch through Every's vibe-check process, where the team tests models against the actual tasks they use daily — shipping features for products like Corora (AI email assistant) and Spyro (AI content automation).

His read on Claude Opus 4 is that it runs autonomously for extended periods on complex pull requests that Claude 3.7 Sonnet couldn't handle, and that it likely exceeds Gemini 2.5 and o3 specifically in Claude Code. On benchmarks, the gains are targeted rather than broad — a meaningful jump on agentic coding, but Claude 3.7 Sonnet actually outperforms Claude Sonnet 4 on visual reasoning by roughly half a percent.

Shipper argues Claude Opus 4 is more directly comparable to Anthropic's hybrid approach — toggling between base-model and reasoning-model behavior — than to GPT-4.5, which is instruction-tuned but not a reasoning model. He frames this hybrid capability as something all model providers are moving toward, but says Anthropic got there first.

On benchmarks generally, Shipper is skeptical: "Most of the benchmarks are kind of bullshit." He points to Llama as a model that scores well on paper but underperforms in practice.

Anthropic's developer positioning

Shipper sees Anthropic consolidating around the developer market rather than competing everywhere. Claude 3.5 Sonnet and 3.7 Sonnet are, in his view, the strongest coding models available, and Claude Code — a command-line interface designed for direct agent interaction rather than an IDE with AI bolted on — is a meaningfully different product from Cursor or Windsurf.

He expects model-switching behavior to slow as intelligence gaps narrow. Developers will increasingly stay inside chosen ecosystems rather than jumping with every release, which benefits incumbents with strong developer workflows.

On image generation, Shipper says he simply doesn't know why Anthropic has no offering, but the absence is conspicuous given Google's momentum with Veo 3. His inference is that Anthropic is deliberately concentrating on agentic coding, where Mike Krieger as CPO gives them product credibility.

Competitive context

Shipper notes he shifted roughly 95% of his personal AI usage to ChatGPT after o3 launched, driven primarily by memory features. He still regards Claude as superior for emotionally nuanced tasks — he uses it to analyze Granola meeting transcripts when navigating management issues — but o3 has become his default for most workflows.

Asked about Google IO, Shipper was at Microsoft Build interviewing CTO Kevin Scott and didn't attend personally, but says Every writer Alex Duffy was visibly moved by what he saw, describing it as "the future" in the team Discord.