How to build the AI platform your company needs
Comments on “How to build the AI platform your company needs”
Claim-checked. Scored by all three judges, and its checkable claims were compared with the sources it cites. How far stories are checked
A long guide uses DoorDash's internal AI platform as its main example for what companies need once several teams build model-powered features: a single gateway to all models, per-team cost tracking, request logs, output safety checks and automated quality tests. DoorDash reports that moving some tasks to open-weight models cut cost about 20x, and that its support-bot checks cut hallucinations by 90%. These figures come from company talks and could not be independently verified. The guide shows other companies making different choices, such as Bitrise building its own gateway.
Why it ranked
No. 12 of 124 with a score of 0.70. The panel rated it highest on practical relevance (0.87) and reader interest (0.80), lowest on significance (0.61). The judges disagreed most on new information (spread 0.46), so the story is marked contested.
Every answer behind this is in the table under Score.
Score
0.70total under rubric v2.2Contestedjudges differ by 0.46 on new information
- Significance0.61
- Reader interest0.80
- Audience breadth0.71
- Practical relevance0.87
- Substantiated0.68
- Technical depth0.73
- New information0.79
- Counts against: Promotional0.12
- Counts against: Overclaim0.22
Each judge’s answer per question (0–1), their mean and spread, and what the question adds to the score (weight × mean).
| Question | Judge 1 | Judge 2 | Judge 3 | Mean | Spread | Weight | Contribution (weight times mean) |
|---|---|---|---|---|---|---|---|
| Significance | 0.72 | 0.40 | 0.70 | 0.61 | 0.32 | +0.24 | +0.146 |
| Reader interest | 0.76 | 0.70 | 0.93 | 0.80 | 0.23 | +0.26 | +0.207 |
| Audience breadth | 0.80 | 0.60 | 0.72 | 0.71 | 0.20 | +0.13 | +0.092 |
| Practical relevance | 0.96 | 0.80 | 0.86 | 0.87 | 0.16 | +0.22 | +0.192 |
| Substantiated | 0.66 | 0.60 | 0.78 | 0.68 | 0.18 | +0.05 | +0.034 |
| Technical depth | 0.78 | 0.60 | 0.80 | 0.73 | 0.20 | +0.05 | +0.036 |
| New information | 0.92 | 0.50 | 0.96 | 0.79 | 0.46 , above the contested threshold | +0.05 | +0.040 |
| Promotional | 0.08 | 0.20 | 0.08 | 0.12 | 0.12 | −0.19 | −0.023 |
| Overclaim | 0.23 | 0.20 | 0.24 | 0.22 | 0.04 | −0.11 | −0.025 |
| Sum of contributions | +0.699 | ||||||
| Claim-check penalty | 0.000 | ||||||
| Total, as published | 0.699 | ||||||
Spread is the highest minus the lowest judge answer. marks a spread above 0.40: the judges disagree, and the story is marked contested.
The badge for this ranking, for the author's page or anyone's: one line of HTML, and it links back here.
<a href="https://rubric.hrih.si/editions/2026-10-27/12/"><img src="https://rubric.hrih.si/badge/2026-10-27/12.svg" alt="Ranked #12 on Rubric, 27 Oct 2026" width="245" height="28"></a>Claim check
Claim check: 2 matches source · 6 couldn’t check. Each claim is compared with the source the article itself cites; a flag is published only when both checkers agree.
Couldn’t checkNumber
DoorDash's internal AI platform serves more than 5,000 internal users, and AI agents call company tools through it 300,000 times a day.
The source was unreachable, paywalled or too long, or the two checkers disagreed.
The article says “their team’s platform serves more than 5,000 internal users. AI agents, programs that act on their own, call company tools through it 300,000 times a day”
The source says No passage could be compared.
Source cited by the article: infoq.com/presentations/doordash-genai-platform-architecture
Couldn’t checkNumber
A tribunal in British Columbia ordered Air Canada to pay a customer C$812 after its chatbot promised a bereavement discount the airline didn't offer.
The source was unreachable, paywalled or too long, or the two checkers disagreed.
The article says “A tribunal in British Columbia ordered Air Canada to pay him C$812, and said why”
The source says No passage could be compared.
Source cited by the article: decisions.civilresolutionbc.ca/crt/crtd/en/item/525448/index.do
Matches sourceNumber
In an a16z survey from January 2026, 81% of enterprises used three or more model families.
The cited source says the same thing.
The article says “81% of the enterprises it asked used three or more model families”
The source says “81% now use three or more model families in testing or production, up from 68% less than a year ago”
Source cited by the article: a16z.com/leaders-gainers-and-unexpected-winners-in-the-enterprise-ai-arms-race
Couldn’t checkNumber
DoorDash moved some tasks to open-weight models such as Qwen hosted on Modal; on those tasks accuracy rose and cost dropped about 20x, with savings in the single-digit millions annually.
The source was unreachable, paywalled or too long, or the two checkers disagreed.
The article says “On those tasks, accuracy went up and “cost drops like 20x”. Chitlur puts the result at “single-digit million-dollar savings annualized”
The source says No passage could be compared.
Source cited by the article: infoq.com/presentations/doordash-genai-platform-architecture
Couldn’t checkNumber
On Vercel's gateway, open-weight models ran 56% of August's tokens but only 14% of spend.
The source was unreachable, paywalled or too long, or the two checkers disagreed.
The article says “On Vercel’s gateway, open-weight models ran 56% of August’s tokens but only 14% of spend”
The source says No passage could be compared.
Source cited by the article: vercel.com/blog/ai-gateway-production-index-september-2026
Couldn’t checkNumber
DoorDash's support bot safety checks cut hallucinations by 90% and severe compliance issues by 99%.
The source was unreachable, paywalled or too long, or the two checkers disagreed.
The article says “Hallucinations, the answers a model makes up, fell by 90%. Severe compliance issues fell by 99%”
The source says No passage could be compared.
Source cited by the article: careersatdoordash.com/blog/large-language-modules-based-dasher-support-automation
Couldn’t checkNumber
A stolen Gemini API key ran up $82,314 in 48 hours for a company of three developers that usually spent $180 a month.
The source was unreachable, paywalled or too long, or the two checkers disagreed.
The article says “a stolen Gemini key that ran up $82,314 in 48 hours. The company, three developers, usually spent $180 a month”
The source says No passage could be compared.
Source cited by the article: theregister.com/2026/03/03/gemini_api_key_82314_dollar_charge
Matches sourceDate
Two poisoned LiteLLM versions were published on March 24, 2026, and FutureSearch counted a 46-minute window and 46,996 downloads.
The cited source says the same thing.
The article says “On March 24, 2026, two poisoned versions of LiteLLM went online”
The source says “During the 46-minute window, the two malicious versions were downloaded 46,996 times”
Source cited by the article: futuresearch.ai/blog/litellm-hack-were-you-one-of-the-47000
Also covered by
Elsewhere
Elsewhere: HN 1 pts, not on front page
Recorded for context only, checked 27 Oct 2026, 05:46 UTC. It played no part in the ranking.