Olivia
Online
Mitchell Agoma
@AgomaMitchell
Building Duetday for couples: one daily question, two private answers, one shared reveal. Private by design. On iPhone ↓
Uk
Joined April 2013
108 Following
117 Followers
4.6K Posts
Pinned Tweet
I built a couples app around a tiny moment most relationship apps ignore: handing one iPhone to your partner and waiting.
One daily question. Two private answers. One shared reveal.
Would your answers match? That’s Duetday. 🧵
@Symbioza2025 @SpaceX @xai @elonmusk @grok Watching the trajectory requires system-level QA: capability growth, tool access, data flows, resource use, human authority and cross-component failure—not model scores in isolation. Related QA view: https://t.co/nhxGFLvPRm
18 AI models. 121 financial-advice questions. A 57% average mistake rate.
That is what Saturn's September 2026 Artificial Authority report says its domain team found across questions on pensions, tax, debt and savings.
The QA lesson is not “AI is useless.”
It is this: fluent output is not release evidence.
For high-risk AI, teams should:
• build domain-expert, risk-based test sets
• repeat the same scenario to expose inconsistency
• verify calculations with deterministic code
• ground answers in current authoritative sources
• test missing context, refusal and escalation
• require human review before consequential action
A green dashboard means little if the oracle is weak.
When one wrong answer could affect a pension, tax decision, debt or savings, “usually helpful” is not a pass criterion.
How would you test an AI adviser before trusting it with a real customer?
Source: https://t.co/Ypm3HBbvOa
#AI #AITesting #QualityEngineering #FinTech #SoftwareTesting
@anishh_3 Strong sequence. I would bring RAG evaluation alongside each stage: parsing loss, chunk recall, retrieval relevance, reranker errors and grounded-answer quality should shape the next design choice. Related QA view: https://t.co/nhxGFLvPRm
18 AI models. 121 financial-advice questions. A 57% average mistake rate.
That is what Saturn's September 2026 Artificial Authority report says its domain team found across questions on pensions, tax, debt and savings.
The QA lesson is not “AI is useless.”
It is this: fluent output is not release evidence.
For high-risk AI, teams should:
• build domain-expert, risk-based test sets
• repeat the same scenario to expose inconsistency
• verify calculations with deterministic code
• ground answers in current authoritative sources
• test missing context, refusal and escalation
• require human review before consequential action
A green dashboard means little if the oracle is weak.
When one wrong answer could affect a pension, tax decision, debt or savings, “usually helpful” is not a pass criterion.
How would you test an AI adviser before trusting it with a real customer?
Source: https://t.co/Ypm3HBbvOa
#AI #AITesting #QualityEngineering #FinTech #SoftwareTesting
@guo_lin99725 Exactly. Evaluator independence requires the right to publish material failures, disclose limitations and reject unsafe release claims. Otherwise “independent” becomes a procurement label. Related QA view: https://t.co/nhxGFLvPRm
18 AI models. 121 financial-advice questions. A 57% average mistake rate.
That is what Saturn's September 2026 Artificial Authority report says its domain team found across questions on pensions, tax, debt and savings.
The QA lesson is not “AI is useless.”
It is this: fluent output is not release evidence.
For high-risk AI, teams should:
• build domain-expert, risk-based test sets
• repeat the same scenario to expose inconsistency
• verify calculations with deterministic code
• ground answers in current authoritative sources
• test missing context, refusal and escalation
• require human review before consequential action
A green dashboard means little if the oracle is weak.
When one wrong answer could affect a pension, tax decision, debt or savings, “usually helpful” is not a pass criterion.
How would you test an AI adviser before trusting it with a real customer?
Source: https://t.co/Ypm3HBbvOa
#AI #AITesting #QualityEngineering #FinTech #SoftwareTesting
@riteshpatel1884 RAG evaluation should come early, not last: retrieval recall, context relevance, answer faithfulness, unsupported claims and failure handling must guide architecture choices. Related QA view: https://t.co/nhxGFLvPRm
18 AI models. 121 financial-advice questions. A 57% average mistake rate.
That is what Saturn's September 2026 Artificial Authority report says its domain team found across questions on pensions, tax, debt and savings.
The QA lesson is not “AI is useless.”
It is this: fluent output is not release evidence.
For high-risk AI, teams should:
• build domain-expert, risk-based test sets
• repeat the same scenario to expose inconsistency
• verify calculations with deterministic code
• ground answers in current authoritative sources
• test missing context, refusal and escalation
• require human review before consequential action
A green dashboard means little if the oracle is weak.
When one wrong answer could affect a pension, tax decision, debt or savings, “usually helpful” is not a pass criterion.
How would you test an AI adviser before trusting it with a real customer?
Source: https://t.co/Ypm3HBbvOa
#AI #AITesting #QualityEngineering #FinTech #SoftwareTesting
@truly_AD The valuable AI-governance professional translates requirements into operating controls—and partners with QA to prove those controls work across roles, data flows, models and exceptions. Related QA view: https://t.co/nhxGFLvPRm
18 AI models. 121 financial-advice questions. A 57% average mistake rate.
That is what Saturn's September 2026 Artificial Authority report says its domain team found across questions on pensions, tax, debt and savings.
The QA lesson is not “AI is useless.”
It is this: fluent output is not release evidence.
For high-risk AI, teams should:
• build domain-expert, risk-based test sets
• repeat the same scenario to expose inconsistency
• verify calculations with deterministic code
• ground answers in current authoritative sources
• test missing context, refusal and escalation
• require human review before consequential action
A green dashboard means little if the oracle is weak.
When one wrong answer could affect a pension, tax decision, debt or savings, “usually helpful” is not a pass criterion.
How would you test an AI adviser before trusting it with a real customer?
Source: https://t.co/Ypm3HBbvOa
#AI #AITesting #QualityEngineering #FinTech #SoftwareTesting
@futuurHQ Different governance models make transparent technical evidence more important. Shared evaluations should document scope, assumptions, failures and uncertainty so policy debate is not driven by slogans alone. Related QA view: https://t.co/nhxGFLvPRm
18 AI models. 121 financial-advice questions. A 57% average mistake rate.
That is what Saturn's September 2026 Artificial Authority report says its domain team found across questions on pensions, tax, debt and savings.
The QA lesson is not “AI is useless.”
It is this: fluent output is not release evidence.
For high-risk AI, teams should:
• build domain-expert, risk-based test sets
• repeat the same scenario to expose inconsistency
• verify calculations with deterministic code
• ground answers in current authoritative sources
• test missing context, refusal and escalation
• require human review before consequential action
A green dashboard means little if the oracle is weak.
When one wrong answer could affect a pension, tax decision, debt or savings, “usually helpful” is not a pass criterion.
How would you test an AI adviser before trusting it with a real customer?
Source: https://t.co/Ypm3HBbvOa
#AI #AITesting #QualityEngineering #FinTech #SoftwareTesting
@nyike The hallucination tax becomes measurable when teams track verification effort, rework, escaped errors and decision delay. Governance should reduce those costs through enforceable quality gates. Related QA view: https://t.co/nhxGFLvPRm
18 AI models. 121 financial-advice questions. A 57% average mistake rate.
That is what Saturn's September 2026 Artificial Authority report says its domain team found across questions on pensions, tax, debt and savings.
The QA lesson is not “AI is useless.”
It is this: fluent output is not release evidence.
For high-risk AI, teams should:
• build domain-expert, risk-based test sets
• repeat the same scenario to expose inconsistency
• verify calculations with deterministic code
• ground answers in current authoritative sources
• test missing context, refusal and escalation
• require human review before consequential action
A green dashboard means little if the oracle is weak.
When one wrong answer could affect a pension, tax decision, debt or savings, “usually helpful” is not a pass criterion.
How would you test an AI adviser before trusting it with a real customer?
Source: https://t.co/Ypm3HBbvOa
#AI #AITesting #QualityEngineering #FinTech #SoftwareTesting
@taiwanplusnews Narrative manipulation is also an evaluation problem. Test source provenance, multilingual consistency, contested claims, refusal behaviour and whether authoritative corrections change the answer. Related QA view: https://t.co/nhxGFLvPRm
18 AI models. 121 financial-advice questions. A 57% average mistake rate.
That is what Saturn's September 2026 Artificial Authority report says its domain team found across questions on pensions, tax, debt and savings.
The QA lesson is not “AI is useless.”
It is this: fluent output is not release evidence.
For high-risk AI, teams should:
• build domain-expert, risk-based test sets
• repeat the same scenario to expose inconsistency
• verify calculations with deterministic code
• ground answers in current authoritative sources
• test missing context, refusal and escalation
• require human review before consequential action
A green dashboard means little if the oracle is weak.
When one wrong answer could affect a pension, tax decision, debt or savings, “usually helpful” is not a pass criterion.
How would you test an AI adviser before trusting it with a real customer?
Source: https://t.co/Ypm3HBbvOa
#AI #AITesting #QualityEngineering #FinTech #SoftwareTesting
@SparkTsaiX Useful distinction. Governance defines authority and constraints; the AI SDLC implements them; QA verifies that controls hold in actual workflows, edge cases and production change. Related QA view: https://t.co/nhxGFLvPRm
18 AI models. 121 financial-advice questions. A 57% average mistake rate.
That is what Saturn's September 2026 Artificial Authority report says its domain team found across questions on pensions, tax, debt and savings.
The QA lesson is not “AI is useless.”
It is this: fluent output is not release evidence.
For high-risk AI, teams should:
• build domain-expert, risk-based test sets
• repeat the same scenario to expose inconsistency
• verify calculations with deterministic code
• ground answers in current authoritative sources
• test missing context, refusal and escalation
• require human review before consequential action
A green dashboard means little if the oracle is weak.
When one wrong answer could affect a pension, tax decision, debt or savings, “usually helpful” is not a pass criterion.
How would you test an AI adviser before trusting it with a real customer?
Source: https://t.co/Ypm3HBbvOa
#AI #AITesting #QualityEngineering #FinTech #SoftwareTesting
@clara_jam_ @nvidia AI regulation debates still need engineering evidence. Teams should be able to show what was tested, under which conditions, which failures remain and who owns the release decision. Related QA view: https://t.co/nhxGFLvPRm
18 AI models. 121 financial-advice questions. A 57% average mistake rate.
That is what Saturn's September 2026 Artificial Authority report says its domain team found across questions on pensions, tax, debt and savings.
The QA lesson is not “AI is useless.”
It is this: fluent output is not release evidence.
For high-risk AI, teams should:
• build domain-expert, risk-based test sets
• repeat the same scenario to expose inconsistency
• verify calculations with deterministic code
• ground answers in current authoritative sources
• test missing context, refusal and escalation
• require human review before consequential action
A green dashboard means little if the oracle is weak.
When one wrong answer could affect a pension, tax decision, debt or savings, “usually helpful” is not a pass criterion.
How would you test an AI adviser before trusting it with a real customer?
Source: https://t.co/Ypm3HBbvOa
#AI #AITesting #QualityEngineering #FinTech #SoftwareTesting
@serlekan An AI security test reaching a real company is a reminder that the harness is part of the system under test. Isolation, synthetic targets, scoped credentials and kill switches are QA requirements. Related QA view: https://t.co/nhxGFLvPRm
18 AI models. 121 financial-advice questions. A 57% average mistake rate.
That is what Saturn's September 2026 Artificial Authority report says its domain team found across questions on pensions, tax, debt and savings.
The QA lesson is not “AI is useless.”
It is this: fluent output is not release evidence.
For high-risk AI, teams should:
• build domain-expert, risk-based test sets
• repeat the same scenario to expose inconsistency
• verify calculations with deterministic code
• ground answers in current authoritative sources
• test missing context, refusal and escalation
• require human review before consequential action
A green dashboard means little if the oracle is weak.
When one wrong answer could affect a pension, tax decision, debt or savings, “usually helpful” is not a pass criterion.
How would you test an AI adviser before trusting it with a real customer?
Source: https://t.co/Ypm3HBbvOa
#AI #AITesting #QualityEngineering #FinTech #SoftwareTesting
@FintechSIN Cross-bank scam models need tests for drift, false positives, adversarial adaptation and unequal impact. Faster detection is useful only with traceable decisions and safe human escalation. Related QA view: https://t.co/nhxGFLvPRm
18 AI models. 121 financial-advice questions. A 57% average mistake rate.
That is what Saturn's September 2026 Artificial Authority report says its domain team found across questions on pensions, tax, debt and savings.
The QA lesson is not “AI is useless.”
It is this: fluent output is not release evidence.
For high-risk AI, teams should:
• build domain-expert, risk-based test sets
• repeat the same scenario to expose inconsistency
• verify calculations with deterministic code
• ground answers in current authoritative sources
• test missing context, refusal and escalation
• require human review before consequential action
A green dashboard means little if the oracle is weak.
When one wrong answer could affect a pension, tax decision, debt or savings, “usually helpful” is not a pass criterion.
How would you test an AI adviser before trusting it with a real customer?
Source: https://t.co/Ypm3HBbvOa
#AI #AITesting #QualityEngineering #FinTech #SoftwareTesting
@riteshpatel1884 That learning path covers the right production layers. The unifying QA skill is turning each architecture promise—retrieval, safety, security and governance—into evidence-backed acceptance criteria. Related QA view: https://t.co/nhxGFLvPRm
18 AI models. 121 financial-advice questions. A 57% average mistake rate.
That is what Saturn's September 2026 Artificial Authority report says its domain team found across questions on pensions, tax, debt and savings.
The QA lesson is not “AI is useless.”
It is this: fluent output is not release evidence.
For high-risk AI, teams should:
• build domain-expert, risk-based test sets
• repeat the same scenario to expose inconsistency
• verify calculations with deterministic code
• ground answers in current authoritative sources
• test missing context, refusal and escalation
• require human review before consequential action
A green dashboard means little if the oracle is weak.
When one wrong answer could affect a pension, tax decision, debt or savings, “usually helpful” is not a pass criterion.
How would you test an AI adviser before trusting it with a real customer?
Source: https://t.co/Ypm3HBbvOa
#AI #AITesting #QualityEngineering #FinTech #SoftwareTesting
@FreedomFireCom @ChuckDBrooks @YuHelenYu @Corix_JC @Shirastweet @Transform_Sec @cybersecboardrm @InfosecurityMag @ScottBVS Business AI governance works when policies become testable controls: approved use cases, data boundaries, evaluation gates, human authority, monitoring and incident ownership. Related QA view: https://t.co/nhxGFLvPRm
18 AI models. 121 financial-advice questions. A 57% average mistake rate.
That is what Saturn's September 2026 Artificial Authority report says its domain team found across questions on pensions, tax, debt and savings.
The QA lesson is not “AI is useless.”
It is this: fluent output is not release evidence.
For high-risk AI, teams should:
• build domain-expert, risk-based test sets
• repeat the same scenario to expose inconsistency
• verify calculations with deterministic code
• ground answers in current authoritative sources
• test missing context, refusal and escalation
• require human review before consequential action
A green dashboard means little if the oracle is weak.
When one wrong answer could affect a pension, tax decision, debt or savings, “usually helpful” is not a pass criterion.
How would you test an AI adviser before trusting it with a real customer?
Source: https://t.co/Ypm3HBbvOa
#AI #AITesting #QualityEngineering #FinTech #SoftwareTesting
@newonpolsia Turning governance into allow, control, escalate or prohibit decisions is practical. QA can make those policies executable by testing each boundary, exception and audit trail. Related QA view: https://t.co/nhxGFLvPRm
18 AI models. 121 financial-advice questions. A 57% average mistake rate.
That is what Saturn's September 2026 Artificial Authority report says its domain team found across questions on pensions, tax, debt and savings.
The QA lesson is not “AI is useless.”
It is this: fluent output is not release evidence.
For high-risk AI, teams should:
• build domain-expert, risk-based test sets
• repeat the same scenario to expose inconsistency
• verify calculations with deterministic code
• ground answers in current authoritative sources
• test missing context, refusal and escalation
• require human review before consequential action
A green dashboard means little if the oracle is weak.
When one wrong answer could affect a pension, tax decision, debt or savings, “usually helpful” is not a pass criterion.
How would you test an AI adviser before trusting it with a real customer?
Source: https://t.co/Ypm3HBbvOa
#AI #AITesting #QualityEngineering #FinTech #SoftwareTesting
@Gartner_inc AI literacy should be demonstrated through role-based capability: recognising limits, verifying evidence, handling sensitive data, escalating risk and understanding who remains accountable. Related QA view: https://t.co/nhxGFLvPRm
18 AI models. 121 financial-advice questions. A 57% average mistake rate.
That is what Saturn's September 2026 Artificial Authority report says its domain team found across questions on pensions, tax, debt and savings.
The QA lesson is not “AI is useless.”
It is this: fluent output is not release evidence.
For high-risk AI, teams should:
• build domain-expert, risk-based test sets
• repeat the same scenario to expose inconsistency
• verify calculations with deterministic code
• ground answers in current authoritative sources
• test missing context, refusal and escalation
• require human review before consequential action
A green dashboard means little if the oracle is weak.
When one wrong answer could affect a pension, tax decision, debt or savings, “usually helpful” is not a pass criterion.
How would you test an AI adviser before trusting it with a real customer?
Source: https://t.co/Ypm3HBbvOa
#AI #AITesting #QualityEngineering #FinTech #SoftwareTesting
@aaakr777 External oversight is valuable when it can inspect evidence and challenge assumptions. AI providers should not be the sole authors, examiners and judges of their own safety claims. Related QA view: https://t.co/nhxGFLvPRm
18 AI models. 121 financial-advice questions. A 57% average mistake rate.
That is what Saturn's September 2026 Artificial Authority report says its domain team found across questions on pensions, tax, debt and savings.
The QA lesson is not “AI is useless.”
It is this: fluent output is not release evidence.
For high-risk AI, teams should:
• build domain-expert, risk-based test sets
• repeat the same scenario to expose inconsistency
• verify calculations with deterministic code
• ground answers in current authoritative sources
• test missing context, refusal and escalation
• require human review before consequential action
A green dashboard means little if the oracle is weak.
When one wrong answer could affect a pension, tax decision, debt or savings, “usually helpful” is not a pass criterion.
How would you test an AI adviser before trusting it with a real customer?
Source: https://t.co/Ypm3HBbvOa
#AI #AITesting #QualityEngineering #FinTech #SoftwareTesting
@MedPoet Subject-matter experts are essential, but their judgments need calibration: clear rubrics, disagreement handling, representative cases and traceability from expert review to model change. Related QA view: https://t.co/nhxGFLvPRm
18 AI models. 121 financial-advice questions. A 57% average mistake rate.
That is what Saturn's September 2026 Artificial Authority report says its domain team found across questions on pensions, tax, debt and savings.
The QA lesson is not “AI is useless.”
It is this: fluent output is not release evidence.
For high-risk AI, teams should:
• build domain-expert, risk-based test sets
• repeat the same scenario to expose inconsistency
• verify calculations with deterministic code
• ground answers in current authoritative sources
• test missing context, refusal and escalation
• require human review before consequential action
A green dashboard means little if the oracle is weak.
When one wrong answer could affect a pension, tax decision, debt or savings, “usually helpful” is not a pass criterion.
How would you test an AI adviser before trusting it with a real customer?
Source: https://t.co/Ypm3HBbvOa
#AI #AITesting #QualityEngineering #FinTech #SoftwareTesting
@oddlab_official Concrete prompts can make model differences visible, but evaluation becomes useful when the expected behaviour, scoring logic, repeatability and consequence of failure are explicit. Related QA view: https://t.co/nhxGFLvPRm
18 AI models. 121 financial-advice questions. A 57% average mistake rate.
That is what Saturn's September 2026 Artificial Authority report says its domain team found across questions on pensions, tax, debt and savings.
The QA lesson is not “AI is useless.”
It is this: fluent output is not release evidence.
For high-risk AI, teams should:
• build domain-expert, risk-based test sets
• repeat the same scenario to expose inconsistency
• verify calculations with deterministic code
• ground answers in current authoritative sources
• test missing context, refusal and escalation
• require human review before consequential action
A green dashboard means little if the oracle is weak.
When one wrong answer could affect a pension, tax decision, debt or savings, “usually helpful” is not a pass criterion.
How would you test an AI adviser before trusting it with a real customer?
Source: https://t.co/Ypm3HBbvOa
#AI #AITesting #QualityEngineering #FinTech #SoftwareTesting
@ForkLog Independent model evaluation earns trust through method, not budget: evaluator autonomy, transparent scope, reproducible failures, published limitations and clear retesting after fixes. Related QA view: https://t.co/nhxGFLvPRm
18 AI models. 121 financial-advice questions. A 57% average mistake rate.
That is what Saturn's September 2026 Artificial Authority report says its domain team found across questions on pensions, tax, debt and savings.
The QA lesson is not “AI is useless.”
It is this: fluent output is not release evidence.
For high-risk AI, teams should:
• build domain-expert, risk-based test sets
• repeat the same scenario to expose inconsistency
• verify calculations with deterministic code
• ground answers in current authoritative sources
• test missing context, refusal and escalation
• require human review before consequential action
A green dashboard means little if the oracle is weak.
When one wrong answer could affect a pension, tax decision, debt or savings, “usually helpful” is not a pass criterion.
How would you test an AI adviser before trusting it with a real customer?
Source: https://t.co/Ypm3HBbvOa
#AI #AITesting #QualityEngineering #FinTech #SoftwareTesting
Last Seen Users on Sotwe
Most Popular Users
Elon Musk
@elonmusk
241.7M followers
Barack Obama
@barackobama
119M followers
Cristiano Ronaldo
@cristiano
114.3M followers
Donald J. Trump
@realdonaldtrump
111.9M followers
Narendra Modi
@narendramodi
107.2M followers
Rihanna
@rihanna
98.7M followers
NASA
@nasa
92.4M followers
Justin Bieber
@justinbieber
91.8M followers
KATY PERRY
@katyperry
90M followers
Taylor Swift
@taylorswift13
83.9M followers
Lady Gaga
@ladygaga
75.4M followers
Virat Kohli
@imvkohli
73.3M followers
Kim Kardashian
@kimkardashian
70.9M followers
YouTube
@youtube
68.8M followers
Neymar Jr
@neymarjr
66.3M followers
Bill Gates
@billgates
65.1M followers
Selena Gomez
@selenagomez
63M followers
The Ellen Show
@theellenshow
62.3M followers
CNN
@cnn
61.8M followers
X
@x
60.7M followers