AI builder & learner | Building awesome stuff + documenting journey to first successful application| DMs open for collabs/advice. Building @Glintbase.
Joined June 2024
- Tweets1.7K
- Following122
- Followers124
- Likes1.8K
Having an llms.txt doesn't make you agent-ready.
Most readiness checks stop at "does the file exist?" Ours asks whether it works.
Here's how the Glintbase score (ARS 3.2) is built 🧵
6/
The part most founders skip:
ARS is not validated as a predictor of agent success.
No study yet links the score to real agent outcomes.
Read it as "readiness signals present and verified," not "agents will succeed." Closing that gap is the actual research.
7/
Full methodology, all 129 checks, and the scanner are open source.
Scan your own site and tell me which check you think is wrong.
Methodology: glintbase.dev/methodology
Blog:glintbase.dev/blog/ars-3-2-e…
Scanner: scan.glintbase.dev
Seriously how do you founders carry out B2B sales?
I've been trying for a while but it seems I'm doing something wrong.
Can I get some guidance pls?
It's actually cause @e2b is agent ready.
I recently ran a study where I benchmarked sandbox providers for agent readiness and E2b was far the easiest for agents to integrate costing far less tokens and number of steps.
Great work from the team optimising their Infra for AI Agents xitter.cf/Connex01/status/210785…
61% of traffic on @e2b is agents.
Agents are already the biggest users of E2B. Soon they'll be also assembling their own infrastructure they need for every task.
We asked 1,773 remote MCP servers to call a tool that doesn't exist.
272 said it worked. HTTP 200, no error signal.
Your agent can't tell the difference either.
Full thread 👇
@dsp_ @RhysSullivan
6/ What we got wrong is in the report too:
• our own pre-registered era-detection rule misread 404/-32601, so we re-ran both rounds
• another crawl overlapped one round, so we discarded it
• our re-probing used up one gateway's free quota
All three are logged as deviations.
7/ 99.6% re-probe agreement, excluding that gateway.
Wilson 95% intervals, host-weighted shares. The harness and per-endpoint data are public.
Full report: glintbase.dev/research/remot…