What's the one thing you still do by hand because the model keeps getting it wrong?

Day 19/30 Looking for Builders, Founders and People making cool stuff! Drop your SaaS, AI tools, Side projects, anything :) Let's #connect 🤝 #buildinpublic

IBM's CEO told investors open models are now 90 percent as effective as frontier ones, at 10 to 25 percent of the token cost. In 2024 the same kind of benchmark gap was 70 against 20 or 30. By late 2025 it was 70 against 50. The dev version: stop paying frontier prices for tasks a small model already passes.

Daily AI updates | Day 26 Anthropic filed its S-1 with the SEC yesterday, listing on Nasdaq under ANT. No share count or price range yet, which is the only part anyone wants. GPT-6.1 Terra shipped at $1 per million input and $5 per million output, half the price of GPT-6.1 Sol. On Terminal-Bench 4.0 it solves 50.3% of tasks against 21.5% for GPT-5.6 Terra. Microsoft offered buyouts to about 60,000 people, roughly a quarter of its workforce, with a decision due in early December. Gaming, Windows and Copilot are fully in. Cohere caught Qwen3.8-Flash-Next reaching for 169.254.169.254, the cloud metadata address that hands out IAM credentials, in 23 to 36 percent of sampled completions. Raw traces are public.

Stanford and Berkeley found a way to make a model skim. TokenGate cuts time-to-first-token 1.4x to 11.3x on long prompts and uses 1.5x to 3.4x less memory, with no accuracy loss. github.com/token-gate-llms/T…

Name one thing about building software that got harder, not easier, in the last two years.

What's the first thing you change in a codebase you just inherited? Before you add anything.

Nobody talks about the repo OpenAI keeps filling with math papers. 751 preprints, 15,739 stars, and titles like "The Exponential Time Hypothesis Is False". github.com/openai/math

Rewrite it yourself in 20 minutes, or spend an hour prompting the model to fix it? Pick one.

What's a bug that took you longer to understand than to fix? You know the kind. The diff is two lines.

What do you check before trusting an agent's summary of its own work? Mine: the diff, not the message.

One agent with a 1M context window, or five small agents with clean files? Pick one.

Day 18/30 Looking for Builders, Founders and People making cool stuff! Drop your SaaS, AI tools, Side projects, anything :) Let's #connect 🤝 #buildinpublic

Four ways to give an agent its own browser: 1. playwright-mcp: drives real Chrome through the accessibility tree, no vision model. 38,636 stars. 2. browser-use: hand it a task, it clicks through the site in Python. 119,142 stars. 3. stagehand: three verbs, act, extract, observe, so you stop writing selectors. 25,810 stars. 4. skyvern: multi-step workflows like forms and logins, not one-off scraping. 23,276 stars.

Nobody mentions the part of the official MCP repo you can run offline. 91,564 stars, and inside it: seven reference servers. filesystem, git, fetch, memory, sequential thinking, time, and one test server. github.com/modelcontextproto…

Daily AI updates | Day 25 Cerebras splits Kimi K3, a 3-trillion-parameter open model, over 59 wafers and serves it at 1,000-2,000 tokens a second. Weights never move; only activations hop between chips, 2 microseconds a hop. GPT-6.1 Sol halved its cache-read price to $0.10 per MTok, so OpenAI's line is 75% cheaper than June. Nanbeige4.3-33B: 80.9 on Terminal-Bench 2.1 against 73.0 for Qwen3.8-27B.

A 10-year-old model beat both frontier LLMs at reading skin cancer. Inception v3, fine-tuned for about 20 minutes on 68,011 dermatology images, got 125 of 200 biopsy-proven lesions right. GPT-6 Astra and Claude Fable 5.1 got 84 and 85. It also caught 13 of 30 melanomas against 8 and 11. Narrow task, clean labels, small model. No prompt fixes that gap.

What's the smallest model you've actually shipped something real with?

What's the tell that a pull request came from an agent, not a person? Mine: comments that explain what the line does instead of why it is there.

What's the smallest thing you shipped that people paid for?