TickerTain
TickerTain
NewsroomShortsPortfolioConvergence
NewsroomShortsPortfolioConvergence
$ANTHROPIC·$MA····$INTC····$BLUE-ORIGIN·$SPCX····$CRWV····$CRM····$MSFT····$NVDA····$ORCL····$CURSOR·$AAPL····$OPENAI·$AMZN····$UBER····$GOOGL····$META····$TSLA····$DATABRICKS·$PERPLEXITY·$LYFT····$NBIS····$TSM····$LITE····$ANDURIL·
$ANTHROPIC·$MA····$INTC····$BLUE-ORIGIN·$SPCX····$CRWV····$CRM····$MSFT····$NVDA····$ORCL····$CURSOR·$AAPL····$OPENAI·$AMZN····$UBER····$GOOGL····$META····$TSLA····$DATABRICKS·$PERPLEXITY·$LYFT····$NBIS····$TSM····$LITE····$ANDURIL·
←

beyang liu

T3 · host / generalist

Co-founder and CTO of Sourcegraph, a developer tools company focused on code search and AI-assisted software development. He was previously an engineer at Palantir.

1 call·1 name·100% bull·last heard 10 months ago·a16z
track record

no scored calls yet — needs a stated position or a categorical verdict, with a matured window vs SPY

top calls

highest conviction · one per company
1sthigh conviction
$SOURCEGRAPHSourcegraphposition

Sourcegraph's AMP agent claims top merge-rate benchmark; pivots to ad-supported free tier

Sourcegraph's coding agent AMP achieved top ranking on a third-party merge-rate benchmark, validating its agent-centric architecture. The company introduced a dual-agent model (smart vs fast) with an ad-supported free tier for the fast agent, expanding TAM beyond enterprise while maintaining usage-based pricing for the premium smart agent.

a16z2025-11episode →

most discussed · click a bar to filter

  • $SOURCEGRAPH

recurring themes

  • AI Coding Agents1
  • AI Infrastructure1
  • Open Source AI1
  • AI Regulation & Policy1
1 total
$SOURCEGRAPH
Sourcegraph
HIGHbeyang liu·a16z·10 months ago·The Truth About Coding Agents: Why 90% of Your Time Is Now Code Review· position
Sourcegraph's AMP agent claims top merge-rate benchmark; pivots to ad-supported free tier
Sourcegraph's coding agent AMP achieved top ranking on a third-party merge-rate benchmark, validating its agent-centric architecture. The company introduced a dual-agent model (smart vs fast) with an ad-supported free tier for the fast agent, expanding TAM beyond enterprise while maintaining usage-based pricing for the premium smart agent.
"Yeah. I think there's like some startup out there that compares uh Polar Quest merge rates or something and we we managed to claim the top spot. ... AMP has two top level agents.…"
5:22
9
AI Coding Agentstailwind
Developers spend 90% of time reviewing agent-generated code; comprehension is new bottleneck
As coding agents generate >90% of code volume, the developer role shifts from writing to orchestrating and reviewing. Current code review tooling (file-by-file diffs) is inadequate for agent-scale changes, creating a product opportunity for agent-native review interfaces that group changes by task and explain intent.
8
AI Infrastructuretailwind
Agent-centric architecture treats model as implementation detail; post-trains small open models for specialized sub-agents
The atomic unit is the agent (model + tools + prompts + environment), not the model. Specialized sub-agents (context retrieval, edit suggestion, debugging) are post-trained from open-weight models at 1B–100B parameter scale, optimizing for latency and task-specific quality rather than general intelligence. This creates a multi-frontier optimization problem per agent.
8
Open Source AIrisk
Chinese open-weight models (Kimi K2, Qwen Coder, GLM) lead in agentic tool use; US open ecosystem lagging
The most capable open-weight models for agentic tool use are now of Chinese origin. Application builders are post-training on these models for specialized sub-agents (search, editing, reasoning), creating a structural dependency risk. US labs (Meta, OpenAI) have not released competitive open-weight models recently, partly due to regulatory/copyright caution.
7
AI Regulation & Policyheadwind
State-by-state AI regulation patchwork entrenches incumbents; federal standards needed for open-weight model competition
Vague, overlapping state regulations on model availability create compliance complexity that only large incumbents can navigate, effectively blocking startups from releasing open-weight models. A clear federal framework targeting applications (not model-layer existential risk) plus enforcement of model-layer competition would preserve a dynamic US open-source ecosystem.